Embed Size (px)
Citation preview
Bulletin
ISSN: 2226-2393
www.icsa.orgJanuary 2014
Volume 26, Issue 1
A Conversation with Dr. Robert O'NeillStatistics for Big Data: Are StatisticiansReady for Big Data?
2013 ICSA AwardsA Snapshot and a Look Ahead ---Members of the ICSA2014 ICSA/KISS Joint SymposiumStatisticians at the United NationsStatisticians at MedImmune Non-ClinicalBiostatistics Group
From the ICSA
meeting at 2013
JSM
ICSABulletin
Volume 26/1, January, 2014ISSN 2226-2393
Editorial StaffEditor-in-Chief
Contributing Editors
Yixuan QiuYihui Xie
Editorial Assistant
Gong-yi Liao
Executive Committee
PresidentYing [email protected]
Past PresidentMing-Hui [email protected]
President-ElectWei [email protected]
Executive DirectorZhezhen [email protected]
TreasurerLinda [email protected]
Consulting EditorFang [email protected] Design: Kan Wu
I'm Just Being Markovian . . . . . . . . . . . . . . . . . . . . . . . 2Late Breaking News . . . . . . . . . . . . . . . . . . . . . . . . . . 3Results of 2013 ICSA Election . . . . . . . . . . . . . . . . . . . . 3From the 2014 President, ICSA . . . . . . . . . . . . . . . . . . . 4From the 2013 President, ICSA . . . . . . . . . . . . . . . . . . . 5From the Executive Director 2014-2016 . . . . . . . . . . . . . . . 7From the Executive Director 2011-2013 . . . . . . . . . . . . . . . 7New Fellows of IMS . . . . . . . . . . . . . . . . . . . . . . . . . . 8People News . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92013 ICSA Awards . . . . . . . . . . . . . . . . . . . . . . . . . . . 10ICSA 2014 Executives and Members of the Committees . . . . . 13A Snapshot and a Look Ahead -- Members of the ICSA . . . . . 15Report of the ICSA Annual Banquet at JSM 2013 . . . . . . . . . 20Report on the 1st ICSA--Canada Chapter Symposium . . . . . . 21Report on ICSA Book Series in Statistics with Springer . . . . . . 23Report From OICSA . . . . . . . . . . . . . . . . . . . . . . . . . . 25Report from the Program Committee . . . . . . . . . . . . . . . . 26Report from ICSA Shanghai Committee . . . . . . . . . . . . . . 28ICSA Financial Report: July 1 through December 31, 2013 . . . 29Report on ICSA Sessions at JSM 2013 . . . . . . . . . . . . . . . . 30Report from the ICSA 2013 International Conference . . . . . . . 31ICSA Sessions for JSM 2014 . . . . . . . . . . . . . . . . . . . . . 322014 ICSA/KISS Joint Symposium Committees . . . . . . . . . . 342014 Student Paper Awards and Travel Grants Announcement . 35Shortcourses at 2014 ICSA/KISS Joint Symposium . . . . . . . . 36A Conversation with Dr. Robert O'Neill . . . . . . . . . . . . . . 42Statisticians at the United Nations . . . . . . . . . . . . . . . . . . 54Statisticians at MedImmune Non-Clinical Biostatistics Group . . 56Statistics for Big Data: Are Statisticians Ready for Big Data? . . . 58Upcoming Events . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Editorial January 2014 Vol.26/1
I'm Just Being MarkovianJun Yan
This issue features two articles. The first is inthe ``Conversation with a Distinguished Statisti-cian'' column, conducted by Dr. Gordan Lan withDr. Robert O'Neill, the Senior Statistical Advisor inthe Center for Drug Evaluation and Research at theFood and Drug Administration (FDA). Dr. O'Neillhas made huge impact on regulatory and pharma-ceutical statistics through both his career at FDAand his visits overseas. He has also been actively in-volved with the growth of ICSA since the very be-ginning, and in 2013, he received an ICSA Distin-guished Achievement Award. The other featuredarticle is in the ``Controversial Issues'' column, onthe very hot topic of big data by Profs. John Jordanand Dennis K.J. Lin. The authors give an overviewof what big data is, how it is done and who areworking with it. Noting the relative absence ofstatisticians in the community, the authors encour-age statisticians to move out of their comfort zone to``be more aggressive, helping or even leading otherdisciplines to advance science and knowledge usingtools that have served us well for centuries, revisedto meet the needs of this new era.''
The ``Statisticians at Work'' column has two ar-ticles, both from statisticians that are working inlesser-known but exciting areas. Dr. Yongyi Minfrom the Statistics Division, United Nations (UN),tells why she chose to work for the UN, what statis-ticians do there, and where they work (literally,all over the world!). Drs. Binbing Yu and HarryYang, Non-Clinical Biostatistics Group at MedIm-mune, tell about what they do and share an exampleof how statisticians bring tremendous value to re-search and development by their diverse skills andexpertise in a collaborative-team environment.
Turning to ICSA business, we have more reportsin this issue than in any other archived issues, in-cluding the first time appearance of reports fromthe ICSA membership committee, the ICSA-CanadaChapter Symposium, the ICSA-Springer Series inStatistics, the ICSA-Shanghai Committee, the ICSAConference in Hong Kong, and ICSA representa-tives at JSM. In particular, Dr. Fang Chen givesa comprehensive statistical analysis with advancedstatistical tools on our members' distribution inspace and growth rate over time.
In 2013, ICSA proudly participated in the cel-ebration of the International Year of Statistics.An article from our Bulletin (Follman, 2013, ICSABulletin, 25:29-30) was used for the ``StatisticianJob of the Week'' on December 20, 2013 (http:
//www.statistics2013.org/2013/12/20/a-statistician-fights-infectious-diseases/).Although 2013 just ended, the celebration goeson. The International Year of Statistics will con-tinue in 2014 and beyond as The World of Statistics.The American Statistical Association (ASA), one ofthe eldest sister associations of ICSA, celebrates its175th Anniversary in 2014. We may as well throwin a celebration for the 55th Anniversary of the Chi-nese Statistical Society (CSS) in the US, the precur-sor of ICSA, established by the Chinese statisticalfaculty and students in Wisconsin (Tiao, Fu, Tsay,and Ting, 2012, ICSA Bulletin, 24:19-22).
Well, sometimes we just celebrate for beingstatisticians. I would like to end my editorial withthe lyrics I used for my daughter's bedtime singing:
You ask me about my occupation.With pride I say that I am a statistician.Like number cruncher, terrible liar, you haveyour expectation.The most I care is about variation.Someone is the best known student.Someone is the most efficient.Someone is always distant.Someone is too busy to get a moment.Willian (Gosset) is the best known student.Emily (MLE) is the most efficient.An outlier is always distant.Cauchy distribution is too busy to get a moment.I like sandwich better than hot dogie,I use it for variance estimation.I like smoothie better than ice cream,it reminds me of spline approximation.Sometimes I feel I have wings of imagination.Those are days I choose to be Bayesian.Sometimes I forget everything but yesterday.That's OK, I'm just being Markovian.
The tunes I used include ``A Girl Worth FightingFor'' (1998 Disney movie Mulan), an antiphonalsinging piece from the 1960 Chinese movie Liu San-jie, and ``One Night to the North of the Great Wall''(2013 Chinese TV series Dragon Gate Express). Youare welcome to try your own tunes.
My best wishes for the coming Spring Festival ofthe Horse Year and a splendid 2014!
Jun YanEditor-in-chief, ICSA BulletinAssociate ProfessorDepartment of StatisticsUniversity of Connecticut
2 © www.icsa.org
January 2014 Vol.26/1 News
Late Breaking NewsStatistica Sinica New Co-Editors
The Statistica Sinica Board of Directors has ap-pointed
• Hsin-Cheng Huang (Academia Sinica)• Ruey Tsay (University of Chicago)• Zhiliang Ying (Columbia University)
as new co-editors and their term is from August 1,2014 to July 31, 2017.
The Memorandum of Understanding(MOU) with JPHCOPH
The MOU between the International ChineseStatistical Association (ICSA) and the Jiann-PingHsu College of Public Health (JPHCOPH) at Geor-gia Southern University (GSU) has been extendedand signed by the ICSA 2013 President and theDean of the JPHCOPH in December 2013. This ex-tension of the MOU covers the academic year pe-riods: 2014-2015, 2015-2016, and 2016-2017, includ-ing summer.
Lingzi Lu Award Committee (ASA/ICSA)The first Lingzi Lu Award Committee has been
appointed jointly by ASA and ICSA in Decem-ber 2013. This committee consists of Ivan Chan(Chair, 2014-2016, ICSA), Gang Li (2014, ICSA),Ying (Grace) Li (2014-2015, ICSA), Eric Kolaczyk(2014, ASA), and Victoria Romberg (2014-2016,
ASA). The committee will select the award recip-ient(s) for attending the Conference on StatisticalPractice.
The First Pao-Lu Hsu AwardThe recipients of the first Pao-Lu Hsu award
were Jianqing Fan, Xiao-Li Meng, and Bin Yu. Theaward ceremony was held at the Ninth ICSA Inter-national Conference on December 20, 2013 at HongKong Baptist University, Hong Kong. The ICSAwould like to thank the Center for Statistical Sci-ence, Peking University (北大统计科学中心) forsponsoring this award and donating the three beau-tiful PL Hsu award plagues.
Lixing Zhu Receives China's State Natu-ral Science Award
Lixing Zhu, Head and Chair Professor of Statis-tics, of the Department of Mathematics, HongKong Baptist University has been bestowed withthe national honour of 2013 State Natural ScienceAward Second Class for his research project enti-tled“Model checking and dimension reduction inregression” . The Natural Science Award honorsachievements in basic research and its applications,covering mathematics, physics, chemistry, astron-omy, geology, biology and other disciplines. It isone of the five categories of China's State Scienceand Technology Awards. In the year of 2013, 53projects in total won this award, 4 of which werein Mathematics.
Results of 2013 ICSA Election2014 President-Elect
Shen, Wei (Eli Lilly and Company)
2014 Biometrics Section Chair
Zeng, Donglin (University of North Carolina atChapel Hill)
Directors of ICSA Board (2014--2016)(alphabetical order)
• Chen, Cong (Merck & Co., Inc)• Chen, Zhen (National Institutes of Health,
USA)• Hsiao, Chuhsing Kate (National Taiwan Uni-
versity)• Jing, Bingyi (Hong Kong University of Science
& Technology)• Liu, Mengling (New York University)• Zhang, Ying (University of Iowa)
© www.icsa.org 3
From the ICSA Executives January 2014 Vol.26/1
From the 2014 President, ICSAYing Lu
Dear ICSA Members andFriends,
Happy New Year to allICSA fellow members andfriends! I would like to thankyou for entrusting me to leadand serve ICSA in 2014. I amhonored to work with you to-gether to continue building
ICSA.In the year 2013, we have achieved historical ac-
complishments. ICSA is a recognized leader amongglobal statistical associations and societies in cele-brating the first International Year of Statistics. Weheld three major statistical conferences in the US,Canada, and Hong Kong; established ICSA firstchapter in Canada; continued scientific leadershipin high quality journal and book publications; ex-tended our collaborations with many statistical as-sociations; and established the Lingzi Lu awardwith the ASA; amongst many other accomplish-ments. These successes were possible because ofthe extraordinary efforts of our dedicated mem-bers. Equally important was the outstanding ICSAleadership of Ming-Hui Chen, ICSA 2013 President,Ivan S. F. Chan, ICSA 2012 President, and Shu-yenHo, the ICSA Executive Director (2011-2013), LindaYau, the ICSA Treasurer (2013-2015), and the Boardof Directors. Their vision, dedication and persis-tent efforts are critical in ICSA's ongoing success.The accomplishments in 2013 has laid solid foun-dation for future growth of the ICSA and set highstandards for 2014.
We are so grateful to have the continuous sup-port from the office of ICSA at the Jian-Ping HsuCollege of Public Health, Georgia Southern Univer-sity for ICSA daily operations, and Lixin (Simon)Gao of BioPier, for the ICSA database and relatedIT operations. There will be a lot of exciting activi-ties in the year of 2014. In June 15-18, the 23nd An-nual ICSA Applied Statistical Symposium will bea joint event with the Korean International Statisti-cal Society (KISS) at Portland Marriott Hotel Water-front, Portland, USA. This will be the first sympo-sium jointly organized by ICSA and KISS. Thanksto Dongseok Choi, Rochelle Fu and Zhezhen Jin fortheir devotion and efforts in organizing this jointsymposium. The first ICSA 2014 Shanghai Statisti-cal Conference will take place in early July in Shang-
hai, China. I would like to thank the ICSA Shang-hai Committee Chair, Dejun Tang and committeemembers, for initiating this event and leading theorganizing activities. These high quality confer-ences will help ICSA to maintain our reputation inscientific leadership. As usual, the ICSA annualmembers meeting will be held at the 2014 JSM inBoston (August 2-7, 2014). I am so grateful for Pro-fessor Tianxi Cai (Chair, the ICSA Annual MeetingCommittee) for her taking on the responsibilities ofplanning and coordinating the ICSA activities at the2014 JSM.
It was a pleasure for me to complete the appoint-ments of various committee chairs and new mem-bers during these last few weeks. ICSA has becomethe fourth largest statistical association in NorthAmerica. We will face new challenges and diffi-cult tasks in maintaining the continuous growthof ICSA. I am fortunate to be able to work withZhezhen Jin, our new executive director (2014-2016). I will work with Ming-Hui Chen, the ICSA2013 President, and Wei Shen, the ICSA 2015 Presi-dent, to develop a strategic vision of ICSA 2020 thatsets the path for future growth.
I will work closely with the ICSA MembershipCommittee chaired by Fang Chen to develop in-novative ideas and strategies to expand the ICSAmembership. As of now, only a small fraction ofICSA members come from Mainland China andEurope. The establishment of new ICSA chaptersacross the globe will expand our membership basegreatly and further increase our international visi-bility. I am very glad to see the progress of ICSAShanghai Committee and looking forward to closercollaborations with statisticians in Mainland China.Ming-Hui and I had a pleasant discussion with Va-lerie Isham, the Past President of the Royal Statis-tical Society (RSS) in Hong Kong. We exchangedgreat ideas to jointly promote our organizations.
In addition to approaching new geographic re-gions, we plan to initiate the use of new technologyand social networking to reach young statisticiansand provide career development assistance. Wemust also reach out to statisticians in different in-dustrial sectors such as finance and insurance com-panies. I will be working closely with the FinanceCommittee chaired by Linda Yau, the Award Com-mittee chaired by Heping Zhang, and the ProgramCommittee chaired by Gang Li (Johnson & Johnson)to examine the feasibility of offering more travelawards to support young statisticians and students,
4 © www.icsa.org
January 2014 Vol.26/1 From the ICSA Executives
especially those from the ICSA priority regions andsectors.
ICSA is unique in its international perspectiveand its culture. I have asked the program committeechaired by Gang Li and the publication committeechaired by Keying Ye to start building a new we-binar program teaching statistics in Chinese. Thisprogram will expand our statistical education andresearch exchange to the larger Chinese speakingcommunity for which English may be a barrier.
ICSA exists because of its members and for itsmembers. ICSA's success depends on you, the vol-unteers and members. I look forward to hearingyour feedback on ICSA activities, criticisms, sugges-
tions, or volunteer offers. Member inputs will keepICSA healthy.
In the year of the horse, I wish you all a happy,healthy, and productive year.
Ying Lu, Ph.D.2014 President, ICSAProfessor of BiostatisticsDepartment of Health Research and PolicyStanford University School of MedicineDirectorPalo Alto VA Cooperative Studies
Program Coordinating CenterPalo Alto VA Healthcare System
From the 2013 President, ICSAMing-Hui Chen
Dear ICSA Members andFriends,
As we start celebrating theyear of the horse on January1, 2014, I have completed theone-year term as the presi-dent of ICSA. I would like totake this opportunity to thankyou for your trust and support
throughout the past year. It has been my great plea-sure to work with so many devoted and dedicatedmembers who have contributed their precious timeand made their efforts in serving this association invarious capacities.
It was the first time in the ICSA history that threemajor ICSA conferences were held in the same year.These include the first ICSA and ISBS joint confer-ence (June 9-12, 2013, the Bethesda North MarriottHotel & Conference Center, Bethesda, Maryland,USA); the first symposium of the ICSA-CanadaChapter (August 2-3, 2013, the Western Harbor Cas-tle Hotel, Toronto, Canada); and the ICSA ninthtriennial international conference (December 20-22,2013, Hong Kong Baptist University, Hong Kong).
More than 1,100 participants attended these con-ferences. It was the first time that the Royal Sta-tistical Society and Hong Kong Statistical Societyco-sponsored the ICSA international conference. Iwould like to thank all of the organizers, programcommittees, and local organizing committees ofthese three conferences for their tremendous effortsto make these events so successful and memorable.
The ICSA journals and publications continue tobe successful in 2013. The term of the current ed-itorial team (Jeng-Min Chiou, Naisyin Wang, andQiwei Yao) for Statisica Sinica (SS) will end in July2014. Hsin-Cheng Huang, Ruey Tsay, Zhiliang Yingwill form a new SS editorial team beginning August2014. During the 2013 JSM, we, including JiahuaChen (Editor-in-Chief of the Springer ICSA BookSeries in Statistics), Shu-yen Ho (ICSA Executive Di-rector), and myself, had a very productive meetingwith the Springer representatives. Several initia-tives had been discussed in promoting this youngbook series. First, the proceedings of ICSA annualsymposium and triennial international conferencewill be published in this book series. Second, threeyears free ICSA membership will be offered to ev-ery book author. Third, handbooks and Springer-Briefs (less than 100 pages) will be included in this
© www.icsa.org 5
From the ICSA Executives January 2014 Vol.26/1
book series. The first proceeding of the ICSA sym-posium, entitled “Topics in Applied Statistics ---2012 Symposium of the International Chinese Sta-tistical Association”, was published by Springer in2013. I would like to thank the editors Mingxiu Hu,Yi Liu and Jianchang Lin for their excellent job inediting this book. The 2013 ICSA symposium bookco-edited by Zhen Chen, Yongming Qu, LianshengLarry Tang, Naitee Ting, and Yi Tsong is expected tobe published in 2014. The organizing committee ofthe ninth ICSA international conference has startedthe preparation of the ICSA international confer-ence book.
As mentioned in my report published in theJuly 2013 issue of the ICSA Bulletin, the ASA andICSA have established a new award in the memoryof Lingzi Lu. I am pleased to report that the firstLingzi Lu Award Committee has been appointedjointly by ASA and ICSA in December 2013. Thiscommittee will select the award recipient(s) to travelto the ASA's annual conference on statistical prac-tice. In order to ensure the success of this award, Isincerely pledge our members to continue to makeyour contributions to this fund. To make your do-nation, please visit: https://www.amstat.org/giving/index.cfm?fuseaction=ShowApp.
When I took the ICSA administrative dutiesfrom Ivan at the beginning of 2013, I had been ac-tively seeking a possibility to establish the ICSA-China Chapter. Many thanks to Dejun Tang (No-vartis, China) for his initiative and tremendousamount of work to prepare the proposal for estab-lishing ICSA Shanghai. The ICSA Board of Direc-tors approved this proposal during the board meet-ing in the 2013 JSM. However, the application pro-cess to establish such an official chapter in Shang-hai is quite complex and the application time couldbe long. Following the IMS China model, the ICSAExecutives decided to establish the Committee forICSA Shanghai as the first step towards the ultimateestablishment of an official ICSA Shanghai chapter.In October 2013, I appointed this committee chairedby Dejun Tang, which consists of 25 members fromuniversities or companies across mainland Chinaplus ICSA Ex-Officio. On Christmas evening (De-cember 25, 2013), the first committee meeting washeld in Shanghai. Both Ying Lu (the ICSA 2014 pres-ident) and I attended this meeting. One of the mainoutcomes of this meeting was to plan the first ICSA
conference at Shanghai in the summer of 2014. Thepreparation of this conference is under way. As youmay know, many Chinese universities have beenexpending the statistics program and the statisticsmajor has become one of the most popular and topchoices for many Chinese students. I hope thatICSA Shanghai will play a crucial role in promotingstatistics and biostatistics in China.
We accomplished a lot in the last year. Thanksto all committee chairs and committee members I orprevious presidents appointed. Without your con-tributions and dedication, none of these would beachieved. Some of you may continue to serve onICSA committees and some have completed yourterm. I would like to take this opportunity to wishyou well in your future endeavors. I know thatthere are still ample tasks to accomplish in the fu-ture. I am sure that President Ying Lu, President-Elect Wei Shen, and the new ICSA Executive Direc-tor Zhezhen Jin will lead the ICSA to the next ad-vancement. The future of ICSA is bright, and ICSAwill continue to grow.
I would like to end this letter by thanking Shu-yen Ho. Shu-yen ended his three-year term as theICSA ED. Shu-yen has made tremendous contribu-tions to ICSA during the last three years. I partic-ularly like Shu-yen's unique skills and approach ineffectively communicating with ICSA board mem-bers, executive members, committee members, andICSA members, as well as in precisely understand-ing and implementing the ICSA By-Laws. I feel sofortunate to have had the opportunity to work withhim last year, and he made my job as the ICSA pres-ident so much easier. I wish Shu-yen well in allhis future ventures. Finally, I would like to thankthe ICSA treasurer Linda Yau for leading the ICSAfinancial team and for taking care of tax-filing toensure the ICSA non-profit status and the OICSAstaffs for working with me last year and for carry-ing out all ICSA daily operations.
I wish you all a happy, healthy, and prosperousyear of the horse.
Ming-Hui Chen, Ph.D.2013 President, ICSAProfessorDepartment of StatisticsUniversity of Connecticut
6 © www.icsa.org
January 2014 Vol.26/1 From the ICSA Executives
From the Executive Director 2014-2016Zhezhen Jin
Dear ICSA members,I am excited and honored
by the opportunity of serv-ing on the role of ExecutiveDirector for next three years(2014--2016). As the ExecutiveDirector, I will work closelywith the President of ICSA,the Board of Directors, the
working committees and the main office to providetimely services to our members.
My predecessor, Dr. Shuyen Ho, has done anexcellent job over the past three years. We haveall witnessed strong growth of our association asa world leading statistical association. Under histremendous support, we have revised our consti-tution and bylaws, improved IT Infrastructure, es-tablished P.L. Hsu Award, established Lingzi LuMemorial Award (joint with ASA), started to estab-lish local chapters, started to publish new SpringerICSA book series, and accomplished many otherachievements.
With all these in mind, I cannot stop think-ing what else we could do to further improve andstrengthen our association. The core of any associ-ation is membership, and ICSA is no exception. We
need to be vigilant and creative in the ways to re-tain current members and to recruit new members.In particular, we need to attract members from stu-dents and from industry and government sectors.It is also critical to identify general interests of ourmembers, and provide our members for their over-all membership benefits. In addition, with emerg-ing new technologies, we need to continue to mod-ernize our IT infrastructure to effectively deliverservices to our members, to facilitate communica-tions among our members, and to increase the vis-ibility of our association. Furthermore, we need tocontinue to work with other statistical societies andorganizations to promote our association.
After all, your support and participation are vi-tal for the continuing success of ICSA. Please let meknow if you have any ideas or suggestions. I lookforward to serving you during next three years, andam optimistic on the future of ICSA with our col-lective efforts and dedications. I wish all of you ahealthy and productive new year.
Zhezhen Jin, Ph.D.ICSA Executive Director (2014-2016)Associate Professor of BiostatisticsDepartment of BiostatisticsColumbia University
From the Executive Director 2011-2013Shuyen Ho
Dear ICSA Members,This is the time I say good
bye, as the Executive Direc-tor of ICSA from 2011 through2013. Over the past threeyears, I have enjoyed workingwith dedicated ICSA officers,board members, committees,office staff and many of ourmembers.
I have also enjoyed watching ICSA progressthroughout my tenure, such as the continuous ex-pansion of the annual Applied Symposium and theTriennial International Conference, which have hadmore and more participants, and increased collab-oration with other organizations. ICSA has also
seen high quality publications in her two journalsand the establishment of the P.L. Hsu and LinzgiLu Awards. I was delighted to have contributed tosome specific improvements within ICSA, such asthe revision of the constitution and bylaws, initiat-ing a Canadian chapter, the up-to-date active mem-bership database with weekly renewal reminders,the office staff home-built annual symposium web-sites, and the new open election candidate nomina-tion procedure. These improvements have helpedICSA grow into a larger, more diversified commu-nity and I look forward to its continued growth inthe future.
However, as we look forward to continuedprogress and expansion in the future, there is stillroom for changes and improvements. From myvantage point, I can see that the complete voluntarybasis of staff and services will need to be modified,
© www.icsa.org 7
People News January 2014 Vol.26/1
and the database structure, tools, and maintenancecould be enhanced or rebuilt. I am fully confidentthat with the new ICSA leadership, ICSA will con-tinue to expand and evolve to become a strongerleading statistics society in the world.
Let me finish my term by completing my re-port on the second 2013 ICSA board meeting. Thismeeting was held on August 4, 2013 at the Mon-treal Convention Center during the Joint Statisti-cal Meetings. The board certified the 2013 electionresults and also updated ICSA finance, the ICSA-Springer Book Series, the membership database, the2013 Awards, the 2013 International Conference inHong Kong and the 2014 ICSA/ KISS (Korean In-ternational Statistics Society) Joint Symposium inPortland Oregon. Subsequently the 2013 electionresults were announced and the 2013 Awards were
presented at the ICSA annual Members meeting onAugust 7.
Again I would like to thank the dedicated ICSAofficers, board members, committees and all of ourmembers for making my term as Executive Directorenjoyable and rewarding. I have gained many valu-able experiences while serving as Executive Direc-tor, and I will treasure it for the rest of my life. I askfor your continued support of and participation inICSA programs and activities.
I wish you all a prosperous year of the Horse!
Shuyen Ho, PhD.ICSA Executive Director (2011-2013)Director, Clinical StatisticsGlaxoSmithKline
New Fellows of IMSJianhua Z. Huang from Texas A&M University,for his contributions to the theory, methodol-ogy and practice of non-parametric and semi-parametric methods, longitudinal and functionaldata analysis, and statistical learning.
Hongzhe Li from University of Pennsylvania, forhis contributions to statistics in genomics, and ap-plications to cancer, autoimmune diseases and mi-crobiology.
Faming Liang from Texas A&M University, forhis contributions to Markov chain Monte Carlomethods and their applications to biology, and forhis service to the profession.
Hua Liang from the University of Rochester, forhis outstanding work on semi-parametric models,model selection and statistical methodology, andhis service to the profession.
Dennis K. J. Lin from Pennsylvania State Univer-sity, for his contributions to experimental designand response surface methodology, and for serviceto the profession.
Jianguo (Tony) Sun from University of Missouri,for his contributions to data analysis, especially inthe analysis of interval-censored failure times, dou-bly censored data and panel count data, and for hisservice to the community.
8 © www.icsa.org
January 2014 Vol.26/1 People News
People NewsJiahua Chen Awarded UBC Killam Re-search Fellowship
Jiahua Chen, former ICSA president (2005),Canada Research Chair (Tier I, 2007-2013), Profes-sor at the Department of Statistics, University ofBritish Columbia (UBC), was awarded a UBC Kil-lam Research Fellowship. This prestigious awardwas to support Professor Chen's research during his2013 study leave. The UBC Killam research fellow-ships are open to all full-time faculty members withtenure or tenure-track, who wish to devote full timeto research and study in their field during a recog-nized study leave. The award is based on specialdistinction of intellect with due regard for soundcharacter and personal qualities.
Conference to be Held in Honor of JeffWu's 65th Birthday in July 2014, Yunnan,China
Building Statistical Methodology and Theory2014, July 7-9, 2014, at Huquan Hotel, Mile (弥勒湖泉酒店), Yunnan, China, is to be held at theoccasion of the 65th birthday of Dr. C. F. JeffWu, one of the most celebrated statisticians of ourtime. Dr. Wu is currently a professor in the Schoolof Industrial and Systems Engineering at Geor-
gia Institute of Technology, Atlanta, USA, and heholds the Coca-Cola Chair in Engineering Statis-tics. Dr. Wu's honors include membership on theNational Academy of Engineering (2004), Member(Academician) of Academia Sinica (2000), COPSS(Committee of Presidents of Statistical Societies)Presidents Award in 1987, and Einstein Chair Pro-fessor at Chinese Academy of Sciences in 2011. He isa fellow of the Institute for Operation Research andthe Management Sciences, of the American Societyfor Quality, of the Institute of Mathematical Statis-tics, and of the American Statistical Association.
The conference intends to bring top researcherstogether to interact and present their new researchfindings in areas that have been influenced by Dr.Wu's previous work and related topics. It is alsoan opportunity for top researchers to share and dis-cuss their ideas and visions of emerging problemsin their fields.
© www.icsa.org 9
ICSA Awards January 2014 Vol.26/1
2013 ICSA AwardsDistinguished AchievementAward
In recognition of the distinguished achievement instatistical research and unselfish support of the associ-ation.
Bin Yu Ph.D., Chancellor's Professor, Universityof California -- Berkeley.
Dr. Bin Yu is a Chancellor's Professor in the De-partments of Statistics and Electrical Engineering& Computer Science at UC Berkeley. She chairedthe Statistics Department at Berkeley from 2009 to2012. She is a founding co-director of the MicrosoftLab on Information Technology and Statistics atPeking University. She received her B.S. in math-ematics from Peking University in 1984, was in theS.S. Chern exchange program in math to the US in1984, and received her M.A. and Ph.D. in Statisticsfrom UC Berkeley in 1987 and 1990, respectively.Before coming back to Berkeley as an Assistant Pro-fessor in 1993, she held faculty positions at Univ ofWisconsin--Madison and Yale University. She wason leave from Berkeley at Bell Labs from 1998 to2000.
She has published more than 100 scientific pa-pers in leading journals and conference proceed-ings on statistics, EECS, remote sensing and neu-
roscience. Her publications cover a wide rangeof research on empirical process theory, informa-tion theory (MDL), MCMC methods, signal pro-cessing, machine learning, high dimensional datainference (boosting and Lasso and sparse model-ing in general and spectral clustering), and inter-disciplinary data problems from neuroscience, re-mote sensing, text summarization and bioinformat-ics. She has served on editorial boards includingAnnals of Statistics, Journal of American StatisticalAssociation, and Journal of Machine Learning Re-search.
Dr. Yu is a Fellow of the American Academyof Arts and Sciences. She was a Guggenheim Fel-low and co-recipient of the Best Paper Award ofIEEE Signal Processing Society in 2006, and was theTukey Memorial Lecturer for the Bernoulli Societyin 2012. She is President-elect of the Institute ofMathematical Statistics (IMS) and an elected Fellowof IMS, AAAS, IEEE, and the American StatisticalAssociation. She serves on the Scientific AdvisoryBoard of the Institute for Pure and Applied Mathe-matics (IPAM) and on the Governing Board of Insti-tute for Computational and Experimental Researchin Mathematics (ICERM), the National Academyof Sciences, and chairs the Scientific Committee ofthe Center of Statistical Science at Peking Univer-sity. She previously served as co-chair of the Na-tional Scientific Committee of the Statistical andApplied Mathematical Sciences Institute (SAMSI)and served on the Boards of Mathematical Sciencesand Applications of the National Academy of Sci-ences (BMSA) and Governors of IEEE--IT Society.
Robert O'Neill Ph.D., Senior Statistical Advisor,Center for Drug Evaluation and Research, FDA.
10 © www.icsa.org
January 2014 Vol.26/1 ICSA Awards
Dr. Robert O'Neill is currently the Senior Statis-tical Advisor in the Office of Translational Sciencesin the Center for Drug Evaluation and Research(CDER), Food and Drug Administration. Up untilJune 2011, Dr. O'Neill was the Director of the Of-fice of Biostatistics which provides biostatistical andscientific computational leadership and support toall programs of CDER. Prior to October 1998 hewas Director of the Office of Epidemiology and Bio-statistics, responsible also for the post-market safetysurveillance of new drugs. In 1989-1990, Dr. O'Neillwas a visiting professor at the Department of Re-search, University Medical School, Basel, Switzer-land, where he developed and presented numer-ous lectures and created a course series Topics inTherapy Evaluation and Review (TITER) for Eu-ropean pharmaceutical scientists, which was themodel for the European Course in PharmaceuticalMedicine (ECPM), a degree granting graduate pro-gram. He has been an FDA topic leader for twoInternational Conference on Harmonization (ICH)guidances, ICH E5 on acceptance of foreign data,and ICH E9 on statistical principles for clinical tri-als. He is a fellow of the American Statistical As-sociation (1985) and the Society for Clinical Trials(2013), a member of several professional societies, apast Member of the Board of Directors of the So-ciety for Clinical Trials, the 2002 recipient of theMarvin Zelen Leadership Award in Statistical Sci-ence, and the 2004 Lowell Reed Lecture Awardeefrom the American Public Health Association. Dr.O'Neill has published over 80 papers and presentednumerous talks including several presentations in2002, 2005 and 2008 to the ICSA.
Outstanding Service Award
In recognition and with sincere appreciation for thededicated effort, unselfish support and outstanding ser-vice.
Lynn Kuo Ph.D., Professor, University of Con-necticut
Dr. Lynn Kuo hasbeen the ICSA sympo-sium account treasurerfrom 2006 to 2009 andthe ICSA treasurer from2009 to 2012. Rightnow she maintains theJ. P. Hsu scholarshipfund and serves as anadviser to the treasurer
and symposium treasurer. The P. L. Hsu scholar-ship fund started in 2010, while she was the trea-surer, as a separate part of the main account formore streamline management. She was also the lo-cal treasurer for the 2006 Applied Statistics Sympo-sium held at the University of Connecticut. In addi-tion to working on the finance for registration, shealso worked with the fund raising committee on ob-taining financial supports. In 2010, she also servedas the U.S. treasurer for the 2010 Eighth ICSA In-ternational Conference at Guangzhou, China. Shehandled all the registration finance for the US andEuropean participants and worked with the In-ternational Conference Committee for a successfulconference. In 2012, she started a conversation andproposal to increase the scholarship awards givenin the annual applied statistics symposium. Withthe board's discussion and approval, this has ledto the present much improved policy on the travelawards with six awards (including one J.P. Hsuaward) and reimbursement up to $1000 each.
Yusong Chen Ph.D., Senior Director, Endo Phar-maceuticals.
Dr. Yusong Chen is the Senior Director ofBiostatistics/Programming of Endo Pharmaceuti-cals, Inc. In his 19-year career in pharmaceu-tical industry, Yusong worked in multiple ther-apeutic areas, such as Oncology, Endocrinolo-gy/Metabolism, Urology, Pain/Anesthesia, andClinical Pharmacology. He is interested in theapplications of pharmacogenomics and pharmaco-dynamics modeling and simulations, adaptive de-signs, diagnostic statistics, Bayesian statistics, tech-niques of dealing with missing data, and statisticalquality control. Yusong earned his Ph.D. in statis-tics from Temple University in 1994. Yusong was acommittee member of ICSA Applied Statistics Sym-posium in 2002. From 2003 to 2012 he served in dif-ferent capacities for ICSA, including the treasurerof ICSA Symposium Account, ICSA Main Account,and JP Hsu Memorial Fund Account. Yusong livesin the Greater Philadelphia Area. He and his wifeWenping Zhang have a daughter, Yue Chen, a son,Stone Chen, and a cat, Kite.
President's CitationIn grateful appreciation of the generosity, dedication
and devoted effort for ICSA.
Ruth Whitworth IT Expert and Webmaster,Jiann-Ping Hsu College of Public Health, GeorgiaSouthern University.
© www.icsa.org 11
ICSA Business January 2014 Vol.26/1
Ms. Ruth Whit-worth currently worksat Georgia SouthernUniversity's Jiann-PingHsu College of PublicHealth as an academicprofessional. She is aliaison between the col-lege and IT Services,as well as the collegewebmaster. She hasan MPH in Biostatistics
from Georgia Southern and an MBA from the Uni-versity of Memphis. She has worked with Karl E.Peace on the Biopharmaceutical Applied StatisticsSymposium (BASS) as a webmaster/registrar since2004 and has helped with the International ChineseStatistical Association server since 2007. Ruth en-joys working at the university, as it provides her anopportunity to help people and learn each day.
Simon Gao Statistical Programming Consultant,Co--Founder, BioPier Inc.
Mr. Lixin (Si-mon) Gao is statisti-cal programming con-sultant and co-founderof BioPier Inc. Heis a permanent mem-ber of ICSA and cur-rently serves as an ITdirector (2010-2012 and2013-2015). Since 2006,he has contributed
tremendous efforts to modernize ICSA informationinfrastructure. He established member databaseand server, enabled online membership registra-tion/renewal and online payment. He also builtelectronic election as well as newsletters. Further-more, he helped past five ICSA symposiums to ac-cept online registrations and abstracts, also servingin their local or advisory committee as well.
Lixin Gao received M.S. in Applied Statisticsfrom Worcester Polytechnic Institute. In this career,he served statistical programming consultancy tomany CROs and bio-pharmaceutical companies inthe field of CSR, CDISC and eSubmission. He isa co-founder of BioPier Inc. where he developedClinical Workbench, an online data review system
for statistical reporting and patient profiles. He alsocreated eSubmission toolkit for companies to sub-mit clinical trial data to FDA.
Lixin Gao started his career as a software engi-neer in Xinhua News Agency back in 1990 after hereceived M.S in Theoretical Physics from ShandongUniversity. Later he worked for Nortel Beijing officebefore he came to the US in 1995.
Lixin Gao also co-founded Hand-by-Hand Edu-cation Foundation (hbhef.org). Since 2006, HBHEFhas sponsored over one thousand students in thepoorest counties of China.
Lili Yu Ph.D., Associate Professor, Jiann-PingHsu College of Public Health, Georgia SouthernUniversity.
Dr. Lili Yu is as-sociate professor inthe department of Bio-statistics in the Jiann-Ping Hsu College ofPublic Health (JPH-COPH) at GeorgiaSouthern University.She joined the Jiann-Ping Hsu College ofPublic Health as anassistant professor in
2007, right after she obtained her Ph.D. in Biostatis-tics from the Ohio State University. At the sametime, she serves as the faculty advisor of the officeof ICSA, which is incubated in the JPHCOPH atGeorgia Southern University. In addition to advis-ing students on the office's daily work, she commu-nicates with different committees in ICSA and pro-vides assistance for the local committees for eachICSA symposium as well. She has been a memberin the ICSA Archive committee from 2008 to 2012,and serves as the chair of this committee in 2013.She has been a reviewer for Jiann-Ping Hsu phar-maceutical and Regulatory Sciences Student PaperAward in ICSA symposium for several years. Shealso serves in Student award committee in 2013 IC-SA/ISBS symposium. Lili Yu's area of research ismainly focused on survival analysis. She works onsemiparametric linear models for heteroscedasticsurvival data in various aspects. Recently, she isalso interested in logistic regressions for categoricaldata with missing values.
12 © www.icsa.org
January 2014 Vol.26/1 ICSA Business
ICSA 2014 Executives and Members ofthe CommitteesEXECUTIVESPresident: Ying Lu (2014)Past President: Ming-Hui Chen (2014)President-elect: Wei Shen (2014)Executive Director: Zhezhen Jin (2014-2016)ICSA Treasurer: Linda Yau (2013-2015)Office of ICSA: Lili Yu, Jingxian Cai, Ruth Whit-worth, Karl Peace, Jiann-Ping Hsu College ofPublic Health, Georgia Southern University,[email protected], Phone: (912) 478-1277.
BOARD of DIRECTORSDonglin Zeng (2014, Biometrics Section Represen-tative) Yeh-Fong Chen (2012-2014), Henry Horng-Shing Lu (2012-2014), Ying Wei (2012-2014), HuiZou (2012-2014), Hongtu Zhu (2012-2014), Yuan-Chin Ivan Chang (2013-2015), Haoda Fu (2013-2015), Wenbin Lu (2013-2015), Ming-Dauh Wang(2013-2015), Zhengjun Zhang (2013-2015), CongChen (2014-2016), Zhen Chen (2014-2016), Chuhs-ing Kate Hsiao (2014-2016), Bingyi Jing (2014-2016),Mengling Liu (2014-2016), Ying Zhang (2014-2016).
STANDING COMMITTEES
Program CommitteeGang Li (Chair, 2014, [email protected])Aiyi Liu (2012-2014), Dongseok Choi (2013-2015),Naitee Ting (2014-2016), Lixing Zhu (2012-2014),Wenqing He (2014-2015), Dejun Tang (2014-2016),Qingxia (Cindy) Chen (2013-2015), Lu Tian (2013-2014), Faming Liang (2014-2016), Tianxi Cai (2013-2014)
Term of reference: (1) Recommend conferenceand symposium sites, including candidates for theirChairs. (2) Recommend general policy for all meet-ings, subject to approval by the Board of Directors.(3) Represent ICSA in the JSM Program Committeeand coordinate ICSA activities at the JSM.
Finance CommitteeLinda Yau (Chair 2013-2015, [email protected])Hongliang Shi (2013-2015), Lynn Kuo (2013-2015),Zhezhen Jin (Ex-Officio 2014-2016)
Term of reference: (1) Manage three ICSA bankaccounts (L. Yau, ICSA main account; H. Shi, ICSAApplied Statistics Symposium account; L. Kuo,ICSA J. P. Hsu Memorial Scholarship Fund ac-count). (2) Oversee the budget and to recommendlong-term financial planning and invest the Asso-ciation's assets, subject to approval by the Board ofDirectors. (3) Manage ICSA PayPal account for on-line credit card payment.
Nomination and Election CommitteeOuhong Wang (Chair, 2014, [email protected])Hsien Ming J Hung (2012-2014), Fenghai Duan(2012-2014), Lixing Zhu (2013-2015), Wei Zhang(2013-2015), Ying Yuan (2013-2015), Xikui Wang(2014-2016), Bing-Shun Wang (2014-2016), JennyHuang (2014-2016),
Term of reference: Nominate the candidates forPresident-elect and members of the Board of Direc-tors.
Publication CommitteeKeying Ye (Chair, 2014, [email protected])Liang Li (2012-2014), Frank Liu (2012-2014), Jun Yan(Editor of Bulletin), Xihong Lin (Co-Editor of SIB),Jose C. Pinhero (Co- Editor of SIB), Hongyu Zhao(Co-Editor of SIB), Jeng-Min Chiou (Co-Editor of S.Sinica), Naisyin Wang (Co-Editor of S. Sinica), Qi-wei Yao (Co-Editor of S. Sinica), Hsin-Cheng Huang(Co-Editor of S. Sinica), Ruey Tsay (Co-Editor ofS. Sinica), Zhiliang Ying (Co-Editor of S. Sinica),Zhezhen Jin (Ex-Officio).
Term of reference: Oversee the publication pol-icy of the Association and make recommendationsto the Board of Directors.
Membership CommitteeFang Chen (Chair, 2014, [email protected])Greg Soon (2012-2014), Cong Chen (2012-2014),Xiaojing Wang (2013-2015), Weiying Wang (2013-2015), Caixia Li (2014-2016), Jianxin Pan (2014-2016).
Term of reference: Recruit new members andcontact interested potential individuals and organi-zations.
© www.icsa.org 13
ICSA Business January 2014 Vol.26/1
Awards CommitteeHeping Zhang (Chair, 2014-2016, [email protected])Yi Tsong (2012-2014), Jack Lee (2012-2014), Gor-don Lan (2013-2015), Xuming He (2013-2015), Mei-Chiung Shih (2014-2016).
Term of reference: Accept, evaluate, and recom-mend nominations for ICSA various awards.
ICSA IT CommitteeLixin (Simon) Gao (Chair, 2013-2015,[email protected])Don Sun, Ruth Whitworth, Zhezhen Jin (Ex-Officio).
2014 Applied Statistics Symposium Com-mitteeDongseok Choi (Co-chair, [email protected]) andRochelle Fu (Co-chair, [email protected])
Term of reference: Organize the Applied Statis-tics Symposium, June 15-18, 2014, Portland Mar-riott Downtown Waterfront, 1401 SW Naito Park-way, Portland, Oregon 97201.
Annual Meeting Committee (2014 JSM)Tianxi Cai (Chair, [email protected])
Term of reference: Plan, coordinate and arrangethe August annual meeting in the 2014 Joint Statisti-cal Meetings, August 2 - 7, 2014, at the Boston Con-vention and Exhibition Center, located at 415 Sum-mer Street, Boston, MA 02210.
ICSA Representative to JSM ProgramCommitteeFaming Liang (2015, [email protected]),Qingxia (Cindy) Chen (2014, [email protected]), Tian Lu (2013, [email protected]).
Term of reference: Represent ICSA in the JSMProgram Committee, coordinate ICSA sponsoredand co-sponsored sessions at JSM.
Archive CommitteeLili Yu (Chair, 2014, [email protected])Smiley Cheng, Shein-Chung Chow, Nancy Lo,Yongming Qu.
Term of reference: Plan and implement elec-tronic archive for the Association.
Committee for ICSA ShanghaiDejun Tang (Chair, Novartis China, [email protected], 10/2013-10/2016)Zhi Geng (Peking University, 10/2013 – 10/2016),Chen Yao (Peking University, 10/2013 – 10/2016),Wei Yuan (Renmin University of China, 10/2013– 10/2016), Qihua Wang (Chinese Academy ofSciences, 10/2013 – 10/2016), Guohua Zou (Chi-nese Academy of Sciences, 10/2013 – 10/2016),Hengjian Cui (Capital Normal University, 10/2013– 10/2016), Zhaojun Wang (Nankai University,10/2013 – 10/2016), Xiaolong Pu (East China Nor-mal University, 10/2013 – 10/2016), Ming Zheng(Fudan University, 10/2013 – 10/2016), NaiqingZhao (Fudan University, 10/2013 – 10/2016), YongZhou (Shanghai University of Finance and Eco-nomic, 10/2013 – 10/2016), Dong Han (Shang-hai Jiao Tong University, 10/2013 – 10/2016),Huazhen Lin (Southwestern University of Financeand Economics, 10/2013 – 10/2016), Xueqin Wang(Sun Yat-sen University, 10/2013 – 10/2016), FengChen (Nanjing Medical University, 10/2013 –10/2016), Jielai Xia (The Fourth Military MedicalUniversity, 10/2013 – 10/2016), Roger Qu (Pfizer,China, 10/2013 – 10/2016), Wei Zhang (Boehringer-Ingelheim, China, 10/2013 – 10/2016), Jie Chen(Merck Serono (Beijing) Pharmaceutical R&D Co.,Ltd., 10/2013 – 10/2016), Bill Wang (Merck, China,10/2013 – 10/2016), Yanping Wang (Eli Lilly andCompany, China, 10/2013 – 10/2016), OuhongWang (Amgen, China, 10/2013 – 10/2016), JunShao (University of Wisconsin/East China NormalUniversity, 10/2013 – 10/2016), Dongchu Sun (Uni-versity of Missouri/ East China Normal University,10/2013 – 10/2016), Ming-Hui Chen (Ex-Officio,2013), Shu-Yen Ho (Ex-Officio, 2013), Ying Lu (Ex-Officio, 2014), Zhezhen Jin (Ex-Officio, 2013-2016),Wei Shen (Ex-Officio, 2015).
Term of reference: Initiate and coordinate ICSAactivities in Mainland China.
Lingzi Lu Award Committee (ASA/ICSA)Ivan Chan (Chair, 2014-2016, ICSA, [email protected])Gang Li (2014, ICSA, [email protected]), Ying(Grace) Li (2014-2015, ICSA), Eric Kolaczyk (2014,ASA), Victoria Romberg (2014-2016, ASA).
BIOMETRICS SECTIONDonglin Zeng (Chair, 2014, [email protected])Aiyi Liu (Past Chair, 2013).
14 © www.icsa.org
January 2014 Vol.26/1 ICSA Reports
A Snapshot and a Look Ahead --Members of the ICSAFang Chen
The early history of the ICSA can be traced backto the 1960's, in a modest Wisconsin house wherea young statistician and his wife took up residenceand entertained Chinese and Taiwanese statisticsstudents and their families. That young statisticianis the greatness-bound George C. Tiao; and thesedinner parties gave birth to the Chinese StatisticalSociety in US, the forebear of the ICSA1. Fast for-ward twenty years, with support from the Instituteof Statistical Science, Academia Sinica, the ICSAwas officially founded at the 1987 Joint StatisticalMeetings in San Francisco, CA.
The ICSA, like many professional and scien-tific societies, started small. The founding mem-bers were only a handful, including George Tiao,James Fu, Smiley Cheng, C.P. Han, Porter Chang,Jeff Wu, L. J. Wei, Jia-yun Tsai, Grace Yang, andGordon Lan2. Some of the names you might rec-ognize; and they are authors of books and papersthat you might have studied. With the dedicationof its members, the society started to expand. Thegrowth of the ICSA also coincided with both thematurity of statistics as a career field and the shiftin demography of those who study and practice it.Twenty-some years after its birth, the society stands1,100 strong and is becoming a vibrant internationalorganization that is fully dedicated to the promo-tion of statistical disciplines and the collaborationof personnel and organizations in many aspects ofsociety.
This report takes a simple snapshot of our mem-bers and offers a look into the future. By under-standing where we are and who we are, maybewe, as an organization, can be better informed andmake better decisions on what we want the societyto become in the future.
The report is largely based on the most recentICSA membership database (current as of Novem-ber of 2013). The membership information is mostlyself-reported -- it is what you provided when youfirst joined the organization; and what you mighthave modified and updated later when you re-newed your membership. Overall, the data is up
to date and the information at hand represents themembers at large well. However, the database doescontain some missing data and out-of-date infor-mation3, therefore, like all things in statistics, inter-pretations and conclusions should be drawn witha proper dose of skepticism along with some light-hearted amusement.
Our MembersThe most basic exploratory statistic is where the
members reside. Table 1 breaks down the mem-bers by their country of residency. The four numbercolumns are the counts for permanent members,regular members, student members (who receive adiscount on the annual membership dues), and thesum of the three. The top five countries account forover 97% of members, and about 2% of memberslive in the other 11 countries.
The United States clearly has an overwhelmingnumber of members. This is not surprising be-cause many of our fellow statisticians are studyingor working in this country. The US also has the moststudent members. The ratio of permanent mem-bers to regular members is about 1-to-1 in the USand Canada; but it is significantly higher in Tai-wan, China, and Singapore, with Singapore claim-ing an astonishing 11 permanent members out of atotal number of 12! The large permanent-to-regularmember ratios outside of North America could bethe result of any or all three of the following reasons:(1) low sign-up numbers of new members in theseplaces (0 student members for example); (2) the has-sle of renewing on a yearly basis from overseas toa US-based website; or (3) the relatively low costof the permanent membership fee ($600) comparedto the regular membership dues ($40 per year). Itprobably is more convenient to pay it off at one time,as opposed to renewing it on a yearly basis.
For members who live in the US, Figure 1 showsresidence counts by the states, discretized into sixgroups. The lighter the color, the fewer registeredmembers. Most members reside in a few statesalong the Atlantic Ocean, with a trend of grad-
1For a more vivid description of the early history of the ICSA, see ``A Conversation with George C. Tiao,'' by Daniel Peña and RueyS. Tsay (2010), Statistical Science, Vol. 25, No. 3, 408--428.
2List is provided by George C. Tiao via personal communication.3If your ICSA information might have been obsolete, please take a moment, log in to your account and update.
© www.icsa.org 15
ICSA Reports January 2014 Vol.26/1
Country Permanent Regular Student Total(%)United States 395 383 65 843 (76.36)Taiwan 72 20 0 92 (8.33)China 58 19 0 77 (6.97)Canada 27 19 8 54 (4.89)Singapore 11 1 0 12 (1.09)Others 8 15 3 26∗ (2.36)Total 571 457 76 1104†*Members from countries that include the United Kingdom (7); Australia and Switzerland (4 each); France, Japan, and Neverlands (2each); and Austria, Belgium, Germany, India, and Saudi Arabia (1 each).†Current as of November of 2013. This number changes almost daily due to the constant ins and outs of new members, renewals, anddropouts of ex-members.
Table 1: Membership by Country of Residency.
Members 0 - 0 1 - 2 3 - 10 11 - 29 32 - 109
AL
AK
AZAR
CA CO
CT
DE
FL
GA
HI
ID
IL IN
IA
KSKY
LA
ME
MD
MAMI
MN
MS
MO
MT
NENV
NH
NJ
NM
NY
NC
ND
OH
OK
OR
PA
RI
SC
SD
TN
TX
UT
VT
VA
WA
WV
WI
WY
PR
Figure 1: Members by State in the U.S.
ual decline towards the West that is interruptedonly by Texas, California, and Washington. Manystates, including New Hampshire, Maine, West Vir-ginia, Oklahoma, North and South Dakota, Mon-
tana, Wyoming, and Idaho, don't even see a singleregistered ICSA member. There are certainly manyreputable institutions with bountiful of statisticiansin this large swath of land. Maybe a concerted effort
16 © www.icsa.org
January 2014 Vol.26/1 ICSA Reports
to recruit in these states would yield new membersin a hurry?
Thanks to the 37 members who reside in Texas,the state is able to buck the declining trend of mem-bers west-bound in the US. Texas A&M Univer-sity (9 members), the M.D. Anderson Cancer Cen-ter (7), and various campuses of the University ofTexas system (10) all contribute to this effort. How-ever, by measure of registered members in a sin-gle organization, the undisputed champion belongsto the FDA, which boasts an impressive 33 regis-tered members--almost as many as the entire stateof Texas. Only a handful of states--Maryland (109,also where most branches of the FDA are located),New Jersey (76), Massachusetts (73), California (70),Pennsylvania (56), North Carolina and New York(50 each), Connecticut and Texas (37 each)--havemore members than the FDA. Whatever the recruit-ing magic that the ICSA had been able to create inthe FDA, we should try to learn and replicate else-where.
Two schools, the University of Michigan (13) andthe University of Connecticut (12), take the top twohonors in the race to the highest population of mem-bers in an educational organization. Quite possiblydue in no small part to tireless effort of our currentpresident Ming-Hui Chen to promote the society,UConn features 8 permanent members, out of 12 intotal! Among for-profit organizations, Pfizer (13)and Eli Lilly (10) are membership heavyweights.The company that I work for, SAS Institute Inc.,doesn't lag too far behind with 9 members. Notably,SAS members also include a rather wealthy billion-aire co-founder and CEO of the company, in JimGoodnight4. The inclusion of Jim Goodnight deliv-ers the dubious title of ``highest average wealth permember in an organization'' to the small enclave ofstatisticians residing in the vicinity of Cary, NorthCarolina.
Next we look at trends of membership sign-ups. Figure 2 breaks down the members into apermanent-member group and a regular-membergroup (the latter includes the student group) andplots them against the dates when the members ini-tially registered. The data is truncated at year 2001to provide a better picture of the membership trendin the past decade. The truncated data have 723records, and they are aggregated to the monthlylevel to reduce noise in the graphs. Note in thegraph that the two y-axes have different limits.
The trend for permanent members is stable,a somewhat flat line that indicates no significant
changes in folks deciding to become a permanentmember of the organization. The trend for regu-lar members tells a different story, with a ratherencouraging upswing slope in the last couple ofyears. While the trend is likely real, the amount ofincrease might be influenced by the few outliers inthe data, noticeably April and May of 2013. Thereis a reason for the two large numbers, that is the2013 Joint Statistics Conference (a joint event hostedby the ICSA and the ISBS), which took place inJune. Historically, ICSA-sponsored conferences arethe biggest drivers of increasing membership ev-ery year. These popular conferences attract a largegroup of statisticians every single time. An addedbenefit to all ICSA members is that you receive aregistration discount. In the past, the ICSA has ef-fectively used this as an avenue to attract new mem-bers and increase the overall membership.
You might be curious to the retention rate ofthe conference-driven new members. Do the newmembers mostly use the opportunity to save somedollars and subsequently shy away from the orga-nization after a year and the membership has ex-pired? For an answer, we look to the month ofJune in 2011, two data points marked on the graphs.In that month, the ICSA Applied Statistics Sympo-sium took place in New York City, which attracteda very large group of attendees. Among the atten-dees were an impressive 254 new members. Threeyears later, 47 of them stayed on with the ICSA (12chose to become permanent members, while therest kept their regular status). A 20% retention rateover three years is a remarkable achievement; andthe June 2012 number is by far greater than thoseof most other months, showing the effect the ICSA-sponsored events have on membership counts. Onthe other hand, I agree that more effort should beput in to retain the more than 200 of these memberswho had since left the organization.
We can discount the two April and May num-bers in the graph by 80% and assume only about 35of these members will stay in the long run. Thatleads to a different picture of the spline fit with asmaller degree of slope. Still, the positive increaseis by no means artificial and I see that as an indicatorfor what is to come in the next few years.
Where Do We Stand?Here are some thoughts on our organization:
• We are doing well in parts of the US, geo-4Jim Goodnight received an honorary permanent membership from the ICSA for his banquet speech in the 2007 Applied Sympo-
sium, Raleigh, North Carolina.
© www.icsa.org 17
ICSA Reports January 2014 Vol.26/1
2013Apr
2013May
2011Jun
0
20
40
60
80
100
Reg
ula
r M
emb
ers
2000 2004 2008 2012
Time
2011Jun
0
10
20
30
40P
erm
enan
t Mem
ber
s
2000 2004 2008 2012
Time
Figure 2: Members' Join Dates. The left panel shows the permanent members; the right panel shows regularmembers. The lines are penalized B-spline fit. Note that the two y-axes do not share the same limit.
graphically and in some industries. There aregaps that could use more filling. Internation-ally, visibility isn't great. We do poorly in at-tracting new members in most places outsideof North America, with very few students oryoung professionals signing up. Historicallyspeaking, statisticians who reside in Chinatend to have a low participation rate, but welook forward to many new fellow members asa result of the recent establishment of ICSAShanghai. We do not yet know what hindersthe growth of the society in places like Tai-wan or Singapore. That is something that themembership committee can explore in the fu-ture.
• A large percentage of the members have re-ceived a Ph.D. or equivalent (over 87% of the
800+ members who had reported their high-est earned degree). There are few master's de-grees, even fewer bachelors. This is somewhatexpected as most Chinese students who cometo the US want to earn an advanced degreeand are often successful in doing so. In addi-tion, most professional conferences and orga-nizations that the ICSA is exposed to or associ-ated with tend to cater to members of this un-usually highly-educated group. On the otherhand, we are missing out on a very large pop-ulation of statisticians by not reaching out tothem. That is something we can think aboutputting a greater effort into. One bright spotis that a high-school graduate joined the ICSAthis year. She had entered an undergraduateprogram in the fall. Young lady, if you readthis article, kudos to you, and we hope that
18 © www.icsa.org
January 2014 Vol.26/1 ICSA Reports
you consider seeking a degree in statistics andjoining the wonderful profession for the manyyears to come!
• The overall membership is steadily growing.The ICSA-sponsored conferences remain thebiggest driver of increasing membership eachyear, and we can successfully retain a portionof these new members over time. The im-provement area is to enable more new mem-bers see beyond the registration-based finan-cial incentive and convince them to stay withthis organization.
Report from the MembershipCommittee
In 2013, the membership committee undertookan effort to clean up the existing ICSA membershipdatabase and sought the renewal of membershipfrom past-due members. The database contains allentries (more than 3,600) that were ever gatheredfrom the ICSA website when members signed up.Over the years, the burden of painstaking mainte-nance of an ever-enlarging and changing member-ship database has fallen upon a few of our fellowmembers. They include Simon Gao (the architec-ture of modern ICSA information infrastructure),Shu-yen Ho (current Executive Director), Ming-HuiChen (past Executive Director and 2013 President),and Lili Yu (of Georgia Southern University). Inspite of their effort, certain individual records re-main out of sync and a number of them are re-dundant -- for example, some members joined thesociety on multiple occasions, creating a new ac-count with each changing work status. This year,the membership committee worked to finish whatthe others have started on this task.
I'd like to acknowledge members of the mem-bership committee for helping out with this impor-tant effort: Zonghui Hu from the National Instituteof Health, Pei Wang from the Fred Hutchinson Can-cer Research Center, Greg Soon from the FDA, CongChen from Merck, Xiaojing Wang from the Univer-sity of Connecticut, and Weiying Wang from M.D.Anderson. We classified members into active, re-dundant, unreachable, and elapsed groups, the lastbeing a collection of slightly over 2,000 ex-members.In September, a renewal reminder email was sent tomembers in the overdue group and asked them to
rejoin the society. We are dispirited to report thatthis effort didn't yield as many renewals as we hadhoped for. It seems that only about 100 or so mem-bers have rejoined. On the bright side, the massivemailing saw a large number of bounced email frominvalid contact addresses, which helped us furtherclean up the membership database.
A Look AheadBy the standard of statistical societies, the ICSA
is a decent-sized organization. Only a few societieshave more members: the ASA (∼19,0005), the IMS(∼4,0006), and ENAR (1,6357). Although the growthpotential is there, the competition is strong. Manyof the societies are looking at the same pool of po-tential members. It is fair to ask how we stand andwhat we can offer to attract new members.
As a society, we do offer good incentives to join.Some of them include free access to some onlinejournals; monthly e-newsletter and biannual ICSAbulletins; the right to vote in ICSA elections and/orbecome a candidate for office; eligibility to serveon ICSA committees and to be involved in ICSA-sponsored or ICSA-cosponsored conferences andworkshops around the world; and last but not least,the potential of networking and establishing profes-sional relationships with different members. In casethe importance of these benefits might not be ap-parent to potential members, we should also thinkof more creative ways to attract young people, stu-dents, professionals who might or might not havea Ph.D., and people who might or might not live inthe US.
One idea that we might want to explore is togradually build the ICSA (especially the ICSA web-site) into a content provider, the go-to place, of awealth of educational materials that offers connec-tions between those who possess in-depth statisticalknowledge with those who seek it. For example, wecan ask members to put together webinars in theirexpertise areas and make them available to all mem-bers online for free. The tutorials need to be wellorganized, generally informative, and taught at in-troductory levels. If someone is looking for a quicklesson on, say, Bayesian computation, adaptive clin-ical trials, or power size calculation, they can accessour server to learn what they need. With a well-organized website that is stocked with good tuto-rials, we can reach people from all across the edu-
5Personal Communication, Bob Rodriguez (2013), ASA president of 20126IMS Wikipedia (2013), http://en.wikipedia.org/wiki/Institute_of_Mathematical_Statistics/7Personal Communication, Challee Blackwelder (2013), Administrator of ENAR
© www.icsa.org 19
ICSA Reports January 2014 Vol.26/1
cational spectrum and geographical locations andmotivate them to join the society. Not only that, theplatform would also provide a place for the lectur-ers to reach a larger professional audience and pos-sibly enhance their own professional development.In addition, accessing ICSA-exclusive material pro-vides a legitimate reason for future members to seekemployer reimbursement for membership dues, aslong as the tutorials offer value for the companiesand the members. The small dues are relatively in-significant for a company but can be a huge incen-tive for attracting new members.
In a finishing thought, the overall health of theICSA is very good, the growth is assured to con-tinue, and the future is rosy. As a member, I wouldlove to see that we continue to expand the organi-zation and attract new members to strive for ourobjectives--that is, to promote statistics and encour-age collaborations among various entities.
Inquires and suggestions are welcome, as al-ways. Please send them [email protected].
AcknowledgmentI would like to thank Ming-Hui Chen and Shu-
yen Ho for their valuable comments and sugges-tions to this manuscript.
Fang Chen, Ph.D.Chair, ICSA Membership Commit-teeSenior ManagerBayesian Statistical ModelingAdvanced Analytics DivisionSAS Institute Inc.
Report of the ICSA Annual Banquet atJSM 2013Xianming Tan
The annual ICSA members meeting was held onAugust 7, 2013, at Westin Hotel, Montreal in Mon-treal, Quebec, Canada. Several awards were givenor announced at the meeting (see photos in this is-sue). After the meeting, about 80 members andguests went to the Le Cristal Chinois (水天一色餐馆), which is located in the heart of the China Townof Montreal, and is about 5 minutes away on footfrom the JSM venue. This restaurant is consideredthe most prestigious chinese restaurant in Montrealand has a solid reputation in this area.
We had a delightful 10-course meal at the ban-quet and here are some of the sample dishes: 乳豬件 (BBQ suckling pork), 薑蔥焗雙龍蝦 (Stir friedlobster with ginger and green onion), 清蒸鱸魚(Steamed fresh bass), 椒牛柳粒蝦球 (Diced Beeftenderloin & shrimp with black pepper). In addi-tion to the wonderful food, the banquet organizersalso arranged for Karaoke. The lighthearted enter-tainment at the end of the evening made the ban-quet ever more fun.
It took quite some effort and time to plan and ar-
range the banquet. The final head count of around80, to the organizers surprise, was quite differentto the predicted headcount of 150--200. However,the banquet went smoothly due to the full supportand prompt decisions of the officials of ICSA: Shu-Yen Ho, Zhezhen Jin, Ming-Hui Chen, Linda Yau,to mention a few.
As an appreciation of the statistics graduate stu-dent volunteers from the McGIll University, Mon-treal, who worked at the ICSA booth during theJSM, the volunteers were treated as guests at thebanquet, in addition to free 1-year ICSA member-ship. The student volunteers were: Zhuoyu Wang,Long Gao, Aixiang Jiang, Zhihui (Amy) Liu, Jun Li,Zhuoyu Sun, Yishu Wang, Jiayi Ni, Menglan Pang.
Xianming Tan, Ph.D.2013 JSM Local Chair, ICSABiostatisticianMcGill University Health CentreEmail: [email protected]
20 © www.icsa.org
January 2014 Vol.26/1 ICSA Reports
Report on the 1st ICSA--Canada ChapterSymposiumGrace Yi
The inaugural symposium of the ICSA-CANADA Chapter has been successfully held atthe Western Harbor Castle Hotel in Toronto onAug. 2-3, 2013. The two-day event was featuredby 12 Special Invited talks, 9 Invited Sessions eachwith 3 talks, 1 session of student contributed talksand a special banquet speech.
We were very proud that this inaugural sympo-sium was held on the International Statistical Year,a special year that marked the unique beginningof the ICSA--CANADA Chapter. The symposiumattracted about 70 faculty members and studentsfrom Canada, the U.S.A. and Taiwan. The attendeesenjoyed the scientific programs which covered a va-riety of emerging issues on statistical sciences andapplications. The symposium apparently provideda friendly venue for people to closely interact bothscientifically and socially.
The Chapter was fortunate to have the presi-dents and past presidents of both ICSA and The Sta-tistical Society of Canada (SSC) to be with us to cel-ebrate its first event. We were honored by the pres-ence of our distinguished speakers and enthusias-tic participants. The participants, regardless beingsenior or junior, offered various kinds of supportto make the symposium a resounding success. Wewould like to take this opportunity to offer a ``bigthank you'' to everybody.
We would like to thank our sponsors, ICSA,Fields Institute and Dalla Lana School of PublicHealth at the University of Toronto for their finan-cial support. We also owe big thanks to the volun-teers for their assistance at the symposium.
The Organizing Committee of the symposiumhad members from both the ICSA-Canada Chap-ter and ICSA. The executive members of the ICSA-Canada Chapter were Jiahua Chen (University ofBritish Columbia), James C. Fu (University of Man-itoba), J. Jack Lee (MD Anderson Cancer Cen-ter), Wendy Lou (University of Toronto) (Co-Chair),and Grace Yi (University of Waterloo) (Co-Chair).The executive members from the ICSA were Jingx-ian Cai (Georgia Southern University), Simon Gao(BioPier Inc.), Shuyen Ho (GlaxoSmithKline Inc.),and Hongliang Shi (Millennium PharmaceuticalsInc.)
A Glimpse of the Symposium
Xiao-Li Meng (Harvard University) kicked offthe symposium with a Special Invited talk of ``Theperils and potential of preprocessing: Explore Dis-tributed Sufficiency'', followed by a Special Invitedtalk given by Michael Evans (University of Toronto)on ``Inferences Derived via a Definition of Statisti-cal Evidence Based on Relative Belief''.
After the morning coffee break, three SpecialInvited talks were delivered. Naisyin Wang (Uni-versity of Michigan) presented a talk of ``Regular-ized Estimation for Ordinary Differential EquationsWith Applications'', Charmaine Dean (University ofWestern Ontario) spoke on ``Joint Analysis of Mul-tivariate Spatial Count and Zero-Heavy Count Out-comes Using Common Spatial Factor Model'', andJane-Ling Wang (University of California, Davis)talked about ``Standard Error Estimation in the EMAlgorithm When Joint Modeling of Survival andLongitudinal Data''.
The afternoon of the first day was divided intotwo periods. The first period was occupied withthree Special Invited talks. Louis-Paul Rivest (Uni-versité Laval) delivered a talk on ``Dealing withHeterogeneity in Capture-Recapture Experiments'',Ming-Hui Chen (University of Connecticut) talkedabout ``Bayesian Inference for Multivariate Meta-Analysis Box-Cox Transformation Models for Indi-vidual Patient Data with Applications to Evalua-tion of Cholesterol Lowering Drugs'', and ChristianLéger (Université de Montréal) concluded this pe-riod with a talk of ``Pseudo-population BootstrapMethods for Imputed Survey Data''.
A second period begun after the afternoon cof-fee break with 4 one-hour invited sessions; two con-current sessions were arranged for each hour. In-vited Session 1 had the theme of ``Recent Advancesin Functional and Dynamic Models'', and had thespeakers of Jiguo Cao (University of Western On-tario), Yichao Wu (North Carolina State University),and Fang Yao (University of Toronto). The secondInvited Session concerned on ``Analysis Methodsfor Complex Structured Data'', and was presentedby Xuewen Lu (University of Calgary), Paul Y. Peng(Queen's University), and Wenqing He (Universityof Western Ontario). Three speakers, Ruodu Wang(University of Waterloo), Min Tsao (University ofVictoria), and Pengfei Li (University of Waterloo)
© www.icsa.org 21
ICSA Reports January 2014 Vol.26/1
spoke at Invited Session 3 on ``Empirical LikelihoodMethods''. Invited Session 4 focused on ``Theoryand Application of Outcome Adaptive Random-ization in Clinical Trials'', and was presented byXikui Wang (University of Manitoba), Yangqing Yi(Memorial University of Newfoundland), and JulieZhou (University of Victoria).
After a long day of scientific sessions, a ban-quet followed at Pearl Harbourfront Chinese Cui-sine in Toronto. Dennis Lin (Penn State University)delivered a speech with the title ``BIGstat@IC2SA''which was informative and entertaining. A nick-name, ``IC2SA'' (pronounced as ``I-C-Square-S-A''),of the ICSA-CANADA Chapter was enjoyed bythe attendees, where the describers ``Chinese'' and``Canada'' were featured with a typical statistical/-mathematical manner: C2.
The second day begun with the Special In-vited talk by Xihong Lin (Harvard School of Pub-lic Health) who presented ``Joint Analysis of SNPand Gene Expression Data in Genetic AssociationStudies of Complex Diseases'', followed by a SpecialInvited talk given by Jack Lee (University of TexasMD Anderson Cancer Center) on ``Bayesian Adap-tive Designs for Efficient Drug Development andImproving Patient Outcomes''. The morning wasconcluded by two Special Invited talks, deliveredby James Fu (University of Manitoba) with a talk on``Finite Markov Chain Imbedding Approximation'',and Jiahua Chen (University of British Columbia)who talked about ``Quantile and Quantile FunctionEstimations under Density Ratio Model''.
The afternoon of the second day was packedup with 5 invited sessions and 1 student con-tributed session; two sessions were concurrentlyscheduled for an hour with one coffee break in-between. Jianan Peng (Acadia University), Wei
Liu (York University), and Lang Wu (University ofBritish Columbia) shared their research on ``Re-cent Developments on Order Restricted Inference''for Invited Session 5. At the concurrent InvitedSession 6, Zeny Feng (University of Guelph), YingZhang (Acadia University) and Xiaoping Shi (Uni-versity of Toronto) spoke on ``Inference Based onNon-parametrics, Pseudo Likelihood or High-orderAsymptotics''. Invited Session 7 had the themeof ``Statistical Inference and Applications''; Yi-TingHwang (National Taipei University), Michelle Liouand Philip Cheng (both from Academia Sinica, Tai-wan) gave talks for this session. Invited Session8 had speakers Wenyu Jiang (Queen's University),Yang Zhao (University of Regina), and Leilei Zeng(University of Waterloo), who discussed ``VariousIssues in the Analysis of Survival and Longitudi-nal Data''. With the theme of ``Modeling and Com-putation for Censored and Spatial Data Analysis'',Invited Session 9 was scheduled for Zhaozhi Fan(Memorial University), Yuehua Wu (York Univer-sity), and Hao Yu (University of Western Ontario).The scientific programs ended with a student con-tributed session of four presentations. Wan-ChenLee (University of Manitoba and University of Win-nipeg), Ying Yan (University of Waterloo), BillyChang (University of Toronto), and Fei Jiang (UTMD Anderson Cancer Center) talked about theirthesis research at this venue.
Grace YiPresident, ICSA--Canada ChapterProfessorUniversity Research ChairDepartment of Statistics and Actu-arial ScienceUniversity of Waterloo
22 © www.icsa.org
January 2014 Vol.26/1 ICSA Reports
Report on ICSA Book Series in Statisticswith SpringerJiahua Chen
The ICSA and Springer have entered a collabo-rative project on the a book series in statistics since2012. The book series aims at publishing bookson different fields related to cutting edge statisti-cal sciences. The series may contain many typesof books such as monographs, contributed works,handbooks, and SpringerBriefs, between 50 and 800pages.
I am honored to be entrusted as the first editorof the book series by the ICSA board through theICSA president Dr. Ivan S. F. Chan in year 2012.In the past year, tens of ICSA members have beencontacted for potential book proposals. Up to thispoint, we have received over five book proposals in-cluding monograph, textbook, and conference sym-posium. The first proposal was from Drs. NaiteeTing and Shuyen Ho. Their book title is ``Statisti-cal Concepts for Health Care Professionals, In Lay-man's terms''. The contract has been signed. Aspromised, Springer has provided a very competi-tive deal. Their book contract is followed by Dr. JingQin, with a book titled ``Empirical likelihood andbiased sampling problems and applications''. It isworthwhile to mention that the research outputs ofDr. Jing Qin form a solid basis or are highly corre-lated to the work by two most recent recipients ofNobel Prize in Economics. I am indeed excited tohave succeeded at persuading Dr. Qin. I anticipatethat this book will be one of the most cited booksin statistics, much like his foundational paper ``Em-pirical likelihood and general estimating equations''(1070 citations so far according to Google Scholar).Due to management team change, this contract isready but it is yet to be signed.
The third proposal comes from ProfessorsChangbao Wu and Mary Thompson. Their book ti-tle is ``Sampling Theory and Practice''. Both Chang-bao and Mary are leading researchers in samplingsurvey. In addition, they have rich experience inteaching and supervising graduate students work-ing in this area. There have been many excellentbooks in sampling survey. However, we feel thatmost of them are either targeting undergraduatestudents or well established practitioners and re-searchers. There is an acute shortage of a bookappropriate for upper year undergraduate and en-trance level graduate students. Their proposal aims
to fill this void.While not finalized yet, Professor Ding-Geng
(Din) Chen has submitted a book proposal on ``Sta-tistical Causal Inferences: Methods and Their Ap-plications'' together with Professors Hua He andPan Wu. To many researchers particularly myself,causal inference is still a foreign notion. I am grate-ful to Professor Din Chen to assemble a team of will-ing experts on the behalf of ICSA Springer book se-ries. Two teams of ICSA members are working onconference proceedings based on ``ICSA 2013 Sym-posium'' and on ``The Ninth ICSA InternationalConference''. Last but not least, I have a book pro-posal myself, which is under review, on theory andapplication of mixture models.
ICSA members form a group of the most activeresearchers and practitioners. It has a huge depositsof book-writing potentials. Yet we are challengedat developing a most effective mining strategy. Of-ten, my inquiry was answered by a regretful replythat a contract with other publishers has only beensigned recently. Whether you are a board mem-ber, member of ICSA, friends of ICSA, please helpto spread the words about the ICSA book series instatistics with Springer. In particular, if you have abook plan, be sure to check with me to turn it into abook-published.
Your participation of this book series is a di-rect support to ICSA. Springer provides competitivedeals to ICSA book writers as well as making a roy-alty contribution to the ICSA. This is a win-win-winproposal. To acknowledge the contributions of thebook writers, the ICSA in addition, pleads 3 yearsfree membership. I sincerely hope to hear more ofyou in the near future.
Jiahua Chen, Ph.D.Executive EditorICSA Springer Series in StatisticsCanada Research Chair(Tier I, 2007-2013)ProfessorDepartment of StatisticsUniversity of British Columbia
© www.icsa.org 23
Start from 2013 Jointly by ICSA & Springer Printed book Hardcover & Softcover
eBook Available from libraries offering Springer’s eBook Collection, or for individual purchase via online bookstores. A free preview is available on SpringerLink. springer.com/ebooks
MyCopy Printed eBook exclusively available to patrons whose library offers Springer’s eBook Collection. springer.com/mycopy
*** Regional restrictions apply.
Kindle edition
Why publish in ICSA and Springer book series?
The International Chinese Statistical Association (ICSA) is among the most influential statistical associations. Its membership currently stands at 1300 (with 530 permanent) and spreads to every corner of the world. Your book will be immediately exposed to a large basis of targeted readers.
The natural connection of the ICSA with China gives your book an edge on the spreading of your academic influence to the largest population and the fastest growing economic power in the world.
Springer is the largest scientific book publisher in the world and has a leading position in electronic publishing. When you publish with Springer your work gets the attention it deserves. Springer values your work and will ensure that your publishing experience remains a personal and individual one.
The cooperation between ICSA and Springer provides an authoritative seal of the quality of your work.
Three years free ICSA membership are offered to every book author.
Editor-in-Chief:
Prof Chen, Jiahua(陈家骅) Department of Statistics, University of British Columbia
Email: [email protected]
Springer Publishing Editor:
Grace Guo(郭海霞), [email protected]
Find Book Manuscript Guidelines and more info on springer.com/authors
January 2014 Vol.26/1 ICSA Reports
Report From OICSALili Yu
Dear ICSA members,Happy New Year! In the past half year, the ICSA
continued making many important and great de-velopments. I would like to take this opportunityto report some important office work and functionsduring the past six months.
The website for ICSA 2014 ICSA symposiumhas been established on the ICSA web, http://www.icsa.org/2014/. The 2014 ICSA sympo-sium will be jointly hosted with the Korean Interna-tional Statistical Society (KISS) at Portland MarriotDowntown Waterfront in Portland, Oregon, June15--18, 2014. The Keynote Speakers and BanquetSpeakers are shown on the website as well as the an-nouncement of the Invited Sessions, Short Coursesand Student Awards. Please visit the website fre-quently to obtain more information about the jointmeeting.
The 9th triennial International Conference ofthe International Chinese Statistical Association(ICSA) was at the Hong Kong Baptist University,Hong Kong from Dec 20--23, 2013. The confer-ence is also co-sponsored by the American Statisti-cal Association (ASA), the Chinese Association ofApplied Statistics (CAAS), the Chinese Society ofProbability and Statistics (CSPS), the Hong KongStatistical Society (HKSS), the Institute of Mathe-matical Statistics (IMS), the International Societyfor Bayesian Analysis (ISBA), and the Royal Sta-tistical Society (RSS). Please refer to the confer-ence websitehttp://www.math.hkbu.edu.hk/ICSA2013/ for more details.
The ICSA Shanghai Committee was establishedin 2013. It is the first ICSA committee establishedin mainland China. The Committee Chair is De-jun Tang who works at Novartis in China. Thecommittee members are composed of statisticians
from both academia and industry in China. It isan important event for the growth of the ICSA,since it increases the connection between statisti-cians in the USA and China. More Statisticians inChina will join ICSA and hence the association willget stronger. More information can be found inthe file --- Committee for ICSA Shanghai at http://www.icsa.org/officers/index.html. Wewill update the website as more information be-comes available.
In order to better enjoy your benefits as mem-bers of ICSA, we remind you to update your in-formation in a timely fashion on the ICSA website.To update your information, please go to the Mem-bers Only Area http://icsa.org/a/member/m_login.jsp. Then use your registered email ad-dress to login to your account. Follow the tab thereif you forgot your password. If you forgot your reg-istered email address, please contact us to obtain.In addition to updating the information, you canaccess two ICSA journals --- Statistica Sinica andStatistics in Biosciences after you login. We hopeall members provide us their correct information,so you can be better served.
To provide excellent service to all members, theoffice of ICSA would like to hear any suggestionsor ideas on improving the work and functionalityof the office. Please feel free to contact and discusswith us any issues you may have related to the officework and functions.
Lili YuOICSAAssociate ProfessorDepartment of BiostatisticsJiann-Ping Hsu College of PublicHealthGeorgia Southern University
© www.icsa.org 25
ICSA Reports January 2014 Vol.26/1
Report from the Program CommitteeZhezhen Jin and Gang Li
ICSA Program Committee• Gang Li
(Chair, 2014)• Tianxi Cai
(2014 JSM local Committee (2014))• Dongseok Choi
(Symposium 2014 (2014--2015))• Wenqing He
(2014-2015)• Faming Liang
(2015 JSM, (2014-2016))• Aiyi Liu
(Symposium 2013 (2014))• Lixing Zhu
(2013 ICSA Conference (2014))• Qingxia (Cindy) Chen
(2014 JSM (2014-2015))• Dejun Tang
(2014-2016)• Lu Tian
(2013 JSM (2014))• Nai-Tee Ting
(Symposium 2015 and ICSA Conference 2016(2014-2017))
Past events since July 20131. ICSA co-sponsored the symposium ``Ad-
vances in Statistical Methods for the Analy-sis of Observational and Experimental Data--- A Symposium in Honor of Anastasios A.(Butch) Tsiatis'' which was held on July 12--13, 2013 in the Department of Statistics, NorthCarolina State University, Raleigh, NC.
2. The first symposium of the ICSA--CanadaChapter was held on August 2-3, 2013 at theWestin Harbour Castle Toronto in Toronto,Canada. The symposium information is avail-able at http://www.icsa.org/CANADA/canada_index.html.
3. The 2013 JSM took place on August 3-8at Montréal, Québec, Canada. It had twoICSA sponsored invited sessions. One wastitled ``Emerging Methodological Issues inPopulation-Based Chronic Disease Research''organized by Professor Jianwen Cai and theother was titled ``Show Case of Analysis ofCorrelated Measurements'' organized by Pro-fessor Naisyin Wang; see a separate report
from Dr. Lu Tian from Stanford University.ICSA member meeting and annual banquetwere held on August 7th. We thank Dr. Xian-ming Tan and his students for their excellentorganization of the event.
4. The 9th triennial ICSA International Confer-ence was held from December 20 to December23, 2013 in at Lam Woo International Confer-ence Centre, Hong Kong Baptist University inHong Kong, China. See conference summaryin a separate report in this bulletin.
Year 20141. The 23rd ICSA 2014 Applied Statistical Sym-
posium will be held from Sunday, June 15 toWednesday, June 18, 2013 at Portland Mar-riott Downtown Waterfront in Portland, OR.It will be a joint conference with the Ko-rean International Society for Statistics (KISS).Drs. Dongseok Choi ([email protected])and Rochelle Fu ([email protected]) co-chairthe event. For details, please see the sep-arate report in the Bulletin and visit ICSAwebsite http://www.statkiss.org/icsakiss2014/.
2. ICSA co-sponsor the symposium ``OrderedData Analysis, Models and health ResearchMethods: An International Conference inHonor of H. N. Nagaraja for His 60th Birth-day'' which will be held on March 7--9, 2014in the Department of Statistics, The Universityof Texas at Dallas, Richardson,TX. The con-ference information is available at http://faculty.smu.edu/ngh/hnnconf.html.
3. ICSA co-sponsor the conference ``The ThirdWorkshop on Biostatistics and Bioinfor-matics'' which will be held on May 9-11,2014 in Georgia State University, Atlanta,GA. The conference information is avail-able at http://www2.gsu.edu/~matyiz/2014workshop/.
4. ICSA co-sponsor the conference ``The 6thRenmin University of China InternationalStatistics Forum'' which will be held on May24-25, 2014 in Renmin University of China,Beijing, China. The conference informationis available at http://stat.ruc.edu.cn/news/tongzhi/2013/1231/674.html.
5. ICSA co-sponsor the conference ``The 11th In-ternational Conference on Ordered Statistical
26 © www.icsa.org
January 2014 Vol.26/1 ICSA Reports
Data'' which will be held on June 2-6, 2014 inMathematical Research and Conference Cen-ter in Bedlewo, Poland. The conference infor-mation is available at http://bcc.impan.pl/14OrderStat/.
6. ICSA co-sponsor the conference ``The ThirdJoint Biostatistics Symposium'' which will beheld on June 27 and 28th, 2014 in Chengdu,Sichuan, China.
7. The 2014 ICSA-China Statistics Conferencewill be held from July 4 to July 5, 2014, in EastChina Normal University, Shanghai, China.This is the first conference organized by theICSA Shanghai committee and will be hostedby East China Normal University.
8. The 2014 JSM will take place at Boston,MA. There will be an ICSA member meet-ing and a banquet. Professor Tianxi Caiat Harvard School of Public Health is theChair for the local committee. Pleasesend your suggestions and questions [email protected].
Year 2015 and beyond1. ICSA 2015 Applied Statistical Symposium
will be held joint with Graybill Confer-ence in Fort Collins, Colorado from Sun-day, June 14 to Wednesday, June 17, 2015.If you would like to help, please contact Dr.Naitee Ting ([email protected]) or program committeeChair Dr. Scott Evans ([email protected]).
2. ICSA 2016 Applied Statistical Symposiumwill be held in Atlanta, GA. ProfessorYichuan Zhao at Georgia State University([email protected]) will be the Chair of or-ganization committee.
3. The 10th ICSA International Conference willtake place in 2016 in Hainan, China. If you
would like to help, please contact Dr. Jun-fang Li ([email protected]) and Dr.Naitee Ting ([email protected]).
4. ICSA 2017 Applied Statistical Symposiumwill be held in Raleigh, NC. Professor Wen-bin Lu at North Carolina State University([email protected]) will be the Chair oforganization committee.
5. ICSA 2018 Applied Statistical Sympo-sium will be held in Piscataway, NJ. Pro-fessor Minge Xie at Rutgers University([email protected]) will be theChair of organization committee.
If you would like to have ICSA co-sponsorshipfor statistical conferences and meetings, please usethe websitehttp://www.icsa.org/meetings/co-sponsorship/index.html to submit yourapplication for co-sponsorship.
If you have comments and suggestions on ICSAprograms, please send your inputs to Dr. Gang Li([email protected]).
Zhezhen Jin, Ph.D.,Chair, ICSA Program Committee(2013)Associate Professor of BiostatisticsDepartment of BiostatisticsColumbia University
Gang Li, Ph.D.,Chair, ICSA Program Committee(2014)Director, Healthcare BioinformaticsMedical Devices & DiagnosticsJohnson & Johnson
© www.icsa.org 27
ICSA Reports January 2014 Vol.26/1
Report from ICSA Shanghai CommitteeDejun Tang
Appointed by the ICSA President, Dr. Ming-Hui Chen, the ICSA Shanghai Committee was es-tablished in October 2013. The committee con-sists of 25 members from academia and industrywho work in China and have expertise in math-ematical statistics and/or different areas of ap-plied statistics, such as engineering, pharmaceu-tical, finance and economics. The members are:Zhi Geng and Chen Yao (Peking University), WeiYuan (Renmin University of China), Qihua Wangand Guohua Zou (Chinese Academy of Sciences),Hengjian Cui (Capital Normal University), ZhaojunWang (Nankai University), Xiaolong Pu, Jun Shao,and Dongchu Sun (East China Normal University),Ming Zheng and Naiqing Zhao (Fudan University),Yong Zhou (Shanghai University of Finance andEconomic), Dong Han (Shanghai Jiao Tong Uni-versity), Huazhen Lin (Southwestern University ofFinance and Economics), Xueqin Wang (Sun Yat-sen University), Feng Chen (Nanjing Medical Uni-versity), Jielai Xia (The Fourth Military MedicalUniversity), Roger Qu (Pfizer, China), Wei Zhang(Boehringer-Ingelheim, China), Jie Chen (MerckSerono (Beijing) Pharmaceutical R&D Co., Ltd), BillWang (Merck, China), Yanping Wang (Eli Lilly andCompany, China), Ouhong Wang (Amgen, China),and Dejun Tang (Novartis, China). The term of thecommittee is 3 years, from Oct 2013 to Oct 2016.In addition, several ICSA Headquarter officers arealso listed as ex-officio for the ICSA Shanghai Com-mittee. They are Ming-Hui Chen (Ex-Officio, 2013),Shu-Yen Ho (Ex-Officio, 2013), Ying Lu (Ex-Officio,2014), Zhezhen Jin (Ex-Officio, 2013-2016), and WeiShen (Ex-Officio, 2015).
In China, there is a large talent pool of statisticalscholars, researchers, analysts, and students. Theestablishment of this committee can further raiseawareness and impact of the ICSA to provide a vi-able venue for statisticians and students in Chinato participate in various ICSA activities. This canstrengthen professional connections and collabora-tions between the ICSA and other statistical orga-nizations in China. It can also host or assist ICSAmeetings and activities in China to provide conve-nience and easy access for most statistical profes-sionals and students in China. Another task of thiscommittee is to assist East China Normal University(ECNU) to establish the ICSA Shanghai Chapter in2014.
We would like to express our sincere appre-ciation to Ming-Hui Chen, Shu-Yen Ho, Ying Lu,Zhezhen Jin, Wei Shen, and the ICSA Board of Di-rectors for their great support to establish the com-mittee. We are very grateful to Jun Shao, XiaolongPu, Qihua Wang, and other committee members inChina for their efforts and hard work during theprocess to create the committee.
After the announcement of the ICSA ShanghaiCommittee, several committee members have al-ready been actively discussing via email or smallgroups during the 9th ICSA International Confer-ence in Hong Kong to organize the inaugural ICSAstatistical conference in China by this committee in2014. Taking advantage of both Ming-Hui Chenand Ying Lu's visiting Shanghai after the 9th ICSAInternational Conference in Hong Kong, we held avery first committee meeting on Dec 25, the Christ-mas day, with the members residing in Shanghai.Ming-Hui Chen, Ying Lu, Jun Shao, Dongchu Sun,Naiqing Zhao, Dong Han, Peng Qu, and Dejun Tangattended the meeting. Yong Zhou sent a representa-tive from his school to join the meeting as well. Themain topic of this meeting was to discuss the overallstrategy and planning for the inaugural ICSA sta-tistical conference in China. The discussions werevery productive and covered almost all major partof the preparations for the conference. It was de-cided that the“2014 ICSA Shanghai Statistical Con-ference”would be tentatively in the first week ofJuly 2014 in Shanghai hosted by ECNU. The ob-jective is to attract domestic researchers and stu-dents to share their work and experience with othercolleagues. Both invited and contributed sessionswill be considered. The committee will work withECNU to plan the conference. Announcementswith further details will be distributed shortly. It isgreatly appreciated that all the ICSA members pro-vide support to this conference.
Dejun Tang, Ph.D.Chair, Committee for ICSA Shang-haiSite HeadBiometrics & Statistical Sciencesand MethodologyIntegrated Information SciencesNovartis Pharma, China
28 © www.icsa.org
January 2014 Vol.26/1 ICSA Reports
ICSA Financial Report: July 1 throughDecember 31, 2013
Beginning Cash Balance 7/1/2013 $145,735.52
Income: $21,446.82
Membership fee $12,860.00
Advertisement fee $100.00
Springer deposit $5,000.00
2013 JSM banquet income $2,554.13
2013 JSM banquet coupon $291.90
Interest Income $640.79
Total income $21,446.82
Expense: $23,121.28
Miscellaneous
PayPal service charge $264.04
2013 JSM member's meeting $323.15
Board meeting (2013 ICSA symposium & JSM) $3,435.80
2013 JSM banquet $3,638.05
2013 JSM banquet misc fees (ticket printing, postage) $57.42
CPA $3,600.00
BioPier, Web Service for 2013 meetings $4,000.00
Virtual storage fee $55.59
TOTAL Miscellaneous $15,374.06
Postage & Delivery
USPS postage $36.41
Foremost $3,715.81
TOTAL Postage & Delivery $3,752.22
Printing and Reproduction
ICSA Bulletin $3,995.00
TOTAL Postage & Reproduction $3,995.00
Total Expense $23,121.28
Net Total Income ‐$1,674.46
Ending Cash Balance 12/31/2013 $144,061.06
International Chinese Statistical Association
Profit and Loss
July 1, 2013 through December 31, 2013
© www.icsa.org 29
ICSA Reports January 2014 Vol.26/1
ASSETS
Checking/Savings
Checking/Savings $144,061.06
TOTAL ASSETS $144,061.06
LIABILITIES & EQUITY
Equity
Opening balance July 1, 2013 $145,735.52
July 1 to December 31, 2013 Net Income ‐$1,674.46
Total Equity $144,061.06
TOTAL LIABILITIES & EQUITY $144,061.06
July 1, 2013 through December 31, 2013
International Chinese Statistical Association
Balance Sheet
Linda Yau, Ph.D.Treasurer (2013--2015), ICSAPrincipal Statistical ScientistBiostatistics, Evidence Generation, US Medical AffairsGenentech
Report on ICSA Sessions at JSM 2013Tian Lu
ICSA has helped the American Statistical Asso-ciation and many other brother associations to orga-nize the Joint Statistical Meeting (2013) successfullyheld at Montreal, Canada. If you have attendedthe meeting, you must have deep impression onthe visibility of ICSA. Especially, ICSA has success-fully sponsored multiple attractive sessions withgood attendance rates. The invitation for sessionproposals received warm responses from members.With a relatively short notice at the end of 2012,we have received more than 20 session proposalscovering broad research areas ranging from classi-cal survival analysis to modern machine learningmethods; from drug development to social networkanalysis; from predicting complex outcome to sci-entific findings in Women's Health Initiative (WHI)study, competing for the two invited sessions. Ithas been an extremely difficult job to find the bestfrom the best! In the end, ICSA selected and spon-sored two invited sessions "Showcase of Analy-
sis of Correlated Measurements" organized by Pro-fessor Naisyin Wang and "Emerging Methodologi-cal Issues in Population-based Chronic Disease Re-search " organized by Professors Charles Kooper-berg and Jianwen Cai. Both sessions received warmresponses from meeting attendees. In addition,ICSA sponsored two topic-contributed sessions andtwo contributed sessions as the primary sponsor.ICSA also co-sponsored other 50+ JSM sessions. Asthe ICSA representative to JSM 2013, I want to ex-press my deepest thanks to all the support receivedfrom ICSA members and wish everyone have agreat 2014
Tian Lu, Ph.D.ICSA Representative, JSM 2013Associate ProfessorDepartment of Health Research andPolicyStanford University School ofMedicine
30 © www.icsa.org
January 2014 Vol.26/1 ICSA Reports
Report from the ICSA 2013 InternationalConferenceSung Nuk Chiu
The ICSA 2013 International Conference wassuccessfully held from Friday December 20 to Mon-day December 23, 2013 at the Hong Kong BaptistUniversity. The theme of the conference was Chal-lenges of Statistical Methods for Interdisciplinary Re-search and Big Data. The total number of attendeeswas 452, coming from 20 countries and regions.
The conference was kicked off with welcomespeeches by Ming-Hui Chen (2013 President, ICSA),Rick Wong (Acting President and Vice-Chancellorof the Hong Kong Baptist University), Way Kuo(President and University Distinguished Professorof City University of Hong Kong), Leslie Tang (Dep-tuy Commissioner for Census and Statistics, HongKong Special Administrative Region), Jeff Wu (Fel-low of the National Academy of Engineering US),Zhiming Ma (Academician of Chinese Academy ofSciences, China), and Ker-Chau Li (Academician ofAcademia Sinica, Taiwan).
Following the opening ceremony was the inau-gural Pao-Lu Hsu Award session, and the first re-cipients of the award, in alphabetical order, wereJianqing Fan (Princeton), Xiao-Li Meng (Harvard)and Bin Yu (UC Berkeley). Jianqing in his talk asked“Are we all wrong?”and we were all relieved when
we realized that Jianqing was not asking whetherthe awards committee was wrong in choosing himas a recipient but asking whether the fundamen-tal assumption in high-dimensional statistics reallyholds and what the consequences are if it does not.Xiao-Li discussed “A trio of inference problemsthat could win you a Nobel prize in Statistics (if youhelp fund it)”. The NP-hard problems (where NPstands for Nobel Prize here) considered are multi-resolution, multi-phase and multi-source inference.Bin talked about“Modeling visual cortex V4 in nat-uralistic conditions with invariant and sparse imagerepresentations”. Not only proposed a new com-putational model for the electrophysiological datacollected, Bin also shared her experience on how tocarry out a successful collaboration with scientists
in other areas.In addition to the talks by the three award
recipients, there were six other plenary talks byRaymond Carroll (Texas A&M), Ching-Shui Cheng(Academia Sinica), Hengjian Cui (Capital Normal),Peter Hall (Melbourne), Tze Leung Lai (Stanford)and Howell Tong (LSE), 248 talks in 74 invitedsessions (including 4 special invited sessions orga-nized by RSS, IMS, CAAS and ISBA and 9 indus-try track sessions) and 109 talks in 21 contributedsessions. Among our speakers we had nine COPSSawardees, as well as a number of academicians ofthe Chinese Academy of Sciences and AcademiaSinica and fellows of the American Academy of Artsand Sciences, Australian Academy of Science, theNational Academy of Engineering US and the RoyalSociety UK. We also had many prominent statisti-cians from industry giving talks and young peoplepresenting their interesting work.
The conference banquet was held in the RoyalPlaza Hotel on Saturday December 21, 2013, wherea booth for fortune telling was set up to offer vari-ous forecasts, probably by some nonlinear models,based on data shown on our palms and faces. It wasa great evening with many laughs and cheers.
In these four days, there were many interestingand exciting scientific discussions inside and out-side the lecture rooms. Some selected pictures areincluded in this issue of the Bulletin and more areavailable at http://www.math.hkbu.edu.hk/ICSA2013/photo/. In conclusion, the conferencewas a huge success.
Sung Nok Chiu, Dr.rer.nat.Chair, Local Organizing Committee2013 ICSA International Confer-ence in HongKongProfessorDepartment of MathematicsHong Kong Baptist University
© www.icsa.org 31
ICSA Reports January 2014 Vol.26/1
ICSA Sessions for JSM 2014Qingxia (Cindy) Chen
Overview The 2014 Joint Statistical Meetings(JSM) will be held August 2-7, at the BostonConvention and Exhibition Center, Boston, Mas-sachusetts. The theme of JSM 2014 is “Statistics:Global Impact – Past, Present, and Future” .This theme, as explained by Jean Opsomer, JSM2014 Program Chair, “seeks to emphasize, cel-ebrate, and share information about the con-tributions our profession has made, currentlymakes, and will continue to make to importantproblems in the world.” The JSM 2014 website,http://www.amstat.org/meetings/jsm/2014/, al-ready contains much information about the confer-ence, and has been and will be updated frequentlyto reflect the work progress of JSM Program Com-mittee. As the ICSA representative to 2014 JSM Pro-gram Committee, I will herein give a summary ofICSA invited sessions for JSM 2014 and would liketo invite proposals for topic-contributed session.
Invited Session Invited Session is the first impor-tant topic that the JSM Program Committees workon. According to the pre-agreements among soci-eties, ICSA has two allocated invited sessions at JSM2014 with one reserved for Statistica Sinica. Duringthe submission period between July 18 and Septem-ber 5, 2013, ICSA received 12 exciting proposals andthey are (ordered by submission date):“EmergingStatistical methods for complex data”(submitted toBiometrics & ICSA), “Latest Advances in Dimen-sion Reduction” , “Superior Small Sample Infer-ences”,“Missing Data in Longitudinal Data Anal-ysis for Clinical Studies”,“Implementation of sta-tistical algorithms in big data platforms”, “NewChallenges in Empirical Likelihood”(submitted toSection on Nonparametrics, the Institute of Math-ematical Statistics, and ICSA), “Statistica SinicaYoung Statisticians Invited Session”(submitted byStatistica Sinica for the reserved slot), “Methodsadjusting for patient switching treatment that dis-rupts randomization in clinical trials” , “Recentadvances in Survival Analysis of Composite End-points”,“Statistical Innovations in Functional Ge-nomics”,“Challenges and Advances in ModelingCancer Surveillance Data” , and “Challenges inModeling Data Complexity”. Choosing one from11 excellent proposals is a hard decision (one pro-posal is reserved for Statistica Sinica slot). Fortu-nately, because this year some entities (societies or
sections) received fewer proposals than they are al-lowed to handle and some have sufficient proposalsfor their allocated slots but not enough to fill com-petition proposals, ICSA, as one of the most com-petitive societies, as well as other entities receiving“surplus”proposals were allowed to swap propos-
als with the entities with insufficient proposals ifthey have proposals that would fit the latter's scope.ICSA sent over some of the excellent proposals toother sections. The proposals“Emerging Statisticalmethods for complex data”,“Latest Advances inDimension Reduction”, and“Challenges in Mod-eling Data Complexity”were therefore sponsoredby Biometrics, Section on Quality & Productivity,and Section on Statistics in Marketing, respectively,as their recommendations for competition propos-als for invited sessions. One of the three proposalswas successfully chosen as invited sessions basedon the average rankings from all of the 2014 JSMProgram Committee Members. In total, 3 out of the12 proposals ICSA received were eventually cho-sen as invited sessions. ICSA will be the primarysponsor for proposals “Implementation of statis-tical algorithms in big data platforms”(organizedby Dr. Henry H.S. Lu and chaired by Dr. Wing H.Wong) and“Statistica Sinica Young Statisticians In-vited Session”(organized and chaired by Dr. QiweiYao), and a co-sponsor for proposal“Emerging Sta-tistical methods for complex data”(organized andchaired by Dr. Lan Xue). The sponsored proposalswill focus on cutting-edge researches conducted onbig data and complex data, including big data plat-forms, detection of genome structural variation, im-pacts of high dimensionality in finite samples, andclassification for complex data, just to name a few.
Topic-contributed (TC) Sessions While the Pro-gram Committee is finalizing the invited program,one of the next upcoming topics is TC sessions. Thisyear JSM has decided to cut back on the number ofTC sessions they allocate to each section. Each sec-tion is given the average number of proposals dur-ing the past three years, reduced by approximately10%. As ICSA sponsored 0, 3, 0, and 2 TC sessions,respectively, during year 2010-2013, ICSA receivesone allocated TC session for JSM 2014. A TC sessionis organized under a common theme and can be inthe form of papers or panels. For a paper basedsession, five participants (5 speakers, or four speak-ers with one discussant, or thee speakers with twodiscussants) are needed; for a panel based session,
32 © www.icsa.org
January 2014 Vol.26/1 ICSA Reports
three to six members are invited to provide com-mentary or a point view on the panel topic. Theadvantages of TC session over contributed sessionsinclude more presentation time (20 minutes versus15 minutes), more cohesive and visible, and largeraudiences. You will need the following informationto submit a TC proposal:
1. Session Type (Topic-Contributed).2. Session Sub Type (Paper or Panel).3. Primary Sponsor (ICSA). A pull-down screen
with a pre-approved list is provided online.4. Session title.5. Session Description. Please include the fol-
lowing information: short description of ses-sion, including focus, content, timeliness andappeal, list of Topic-Contributed speakers/-panelists, tentative title for each presentationand the session format (e.g., chair, 3 speakersand discussant). Individual abstracts will berequired by February 3, 2014, if the proposedsession is approved.
6. Session chair, including affiliation, address,telephone number and e-mail address.
7. Discussant (if any), including affiliation, ad-dress, telephone number and e-mail address.
If you are interested in organizing a TC ses-sion, please visit http://www.amstat.org/meetings/jsm/2014/topiccontributed.cfm. The deadline to submit TC session proposalis January 15, 2014 at 11:59pm ET. I am lookingforward to receiving your exciting proposal.
Poster, SPEED, and Contributed Sessions Fi-nally, I want to remind you the other con-tributed categories including poster, SPEED,
and contributed sessions. Due to lim-ited space, I will refer readers to http://magazine.amstat.org/blog/2013/12/01/contributedabstracts2014/ for more infor-mation on the traditional poster and contributedsessions and take this opportunity to introduce thenew type of contributed session, SPEED. A SPEEDsession, implying by its name, consists of 20 quickoral presentations of five minutes each, with 10minutes break after the first set of 10 talks. The shortoral presentations are then followed by a poster ses-sion on the same day. SPEED was on a trial on JSM2013 as collaborations within a few ASA sectionsand received very positive feedback. Therefore, itwill be implemented in a wider range with doubledvolume of 10 SPEED sessions on JSM 2014. ICSAdidn't participate in the SPEED sessions in 2014, butplease be aware of this new trend and if you could,attend some of the SPEED sessions, so ICSA can or-ganize a SPEED session in the future with enoughinterests from our members.
If you have any questions or comments aboutJSM 2014, please feel free to contact me [email protected]. I will be look-ing forward to seeing you at JSM 2014.
Qingxia (Cindy) Chen, Ph.D.ICSA Representative, JSM 2014Department of BiostatisticsVanderbilt University School [email protected]
© www.icsa.org 33
2014 ICSA Applied Statistics Symposium January 2014 Vol.26/1
2014 ICSA/KISS Joint SymposiumCommitteesExecutive Committee
• Dongseok Choi, Co-Chair, Oregon Health &Science U.
• Rochelle Fu, Co-Chair & Treasurer, OregonHealth & Science U.
• Joan Hu, Simon Fraser U.
• Zhezhen Jin, Program Chair, Columbia U.
• Ouhong Wang, Amgen
• Ru-Fang Yeh, Genentech
• XH Andrew Zhou, U. of Washington
• Cheolwoo Park, Webmaster, U. of Georgia
Local Committee• Rochelle Fu, Chair, Oregon Health & Science
U.• Yiyi Chen, Oregon Health & Science U.• Thuan Nguyen, Oregon Health & Science U.• Byung Park, Oregon Health & Science U.• Xinbo Zhang, Oregon Health & Science U.
Program Committee• Zhezhen Jin, Chair, Columbia U.• Gideon Bahn, VA Hospital• Kani Chen, Hong Kong U. of Science and
Technology• Yang Feng, Columbia U.• Liang Fang, Gilead• Qi Jiang, Amgen• Mikyoung Jun, Texas A&M U.• Sin-Ho Jung, Duke U.• Xiaoping Sylvia Hu, Gene• Jane Paik Kim, Stanford U.• Mimi Kim, Albert Einstein College of
Medicine• Mi-OK Kim, Cincinnati Children's Hospital
Medical Center
• Xue Lan, Oregon State U.• Gang Li, Johnson and Johnson• Yunfeng Li, Phamacyclics• Mei-Ling Ting Lee, U. of Maryland• Yoonkyung Lee, Ohio State U.• Meng-Ling Liu, New York U.• Xinhua Liu, Columbia U.• Xiaolong Luo, Celgene Corporation• Taesung Park, Seoul National U.• Yu Shen, MD Anderson Cancer center• Greg (Guoxing) Soon, U.S. Food and Drug
Administration• Zheng Su, Deerfield Company• Christine Wang, Amgen• Yichuan Zhao, Georgia State U.
Student Paper Award Committee• Wenqing He, Chair, U. of Western Ontario• Qixuan Chen, Columbia U.• Hyunson Cho, National Cancer Institute• Dandan Liu, Vanderbilt U.• Jinchi Lv, U. of Southern California
Short Course Committee• Xiaonan Xue, Chair, Albert Einstein College
of Medicine• Wei-Ting Hwang, U. of Pennsylvania• Ryung Kim, Albert Einstein College of
Medicine• Jessica Kim, U.S. Food and Drug Administra-
tion• Laura Lu, U.S. Food and Drug Administration• Mikyoung Jun, Texas A&M U.• Tao Wang, Albert Einstein College of
Medicine
IT Support• Lixin (Simon) Gao, Biopier Inc.
34 © www.icsa.org
January 2014 Vol.26/1 2014 ICSA Applied Statistics Symposium
2014 Student Paper Awards and TravelGrants Announcement
The 2014 Joint Applied Statistics Symposiumby the International Chinese Statistical Association(ICSA) and the Korean International Statistical So-ciety (KISS) will be held during June 15--18, 2014 atPortland Marriot Downtown Waterfront, Portland,Oregon, This will be the 23rd annual symposiumfor ICSA.
To encourage student members of ICSA andKISS to participate and to share their research at theSymposium, ICSA and KISS will offer Student Pa-per Awards and Travel Grants for outstanding stu-dent papers.
Qualification The applicant must be an ICSA orKISS member at the time of manuscript submission,a doctoral degree candidate in any term during theacademic year of 2013-2014 at an accredited insti-tute, and be able to register and present the researchwork at the 2014 Joint Applied Statics Symposium.
Manuscript Requirement Manuscript should beprepared double spaced using Biometrics or JASAguidelines for authors. Excluding tables and fig-ures, the manuscript must be no more than 20pages, with at least one-inch for all margins andno smaller than 12-point font. The research workmust be relevant to application in a variety of fieldsincluding biomedicine, finance, business, etc. Themanuscript may be co-authored with a faculty ad-visor and/or a small number of collaborators, butthe student must be the leading author.
Submission of Manuscript Manuscript shouldbe received no later than March 15, 2014. The sub-mission should include:
• A cover letter;
• A separate title page with author(s), institu-tional affiliation, mailing address, phone/faxnumbers and email address;
• A separate page of abstract;
• A blind copy of the manuscripts without au-thor information or affiliation;
• A copy of the ICSA membership applica-tion form for non-members. (Membershipapplication/renew forms can be found fromhttp://www.icsa.org.)
All materials should be packaged into one .zipfile and sent by email to ICSA Student Award Com-mittee at [email protected].
Withdraw and Disqualification Applicants arefree to withdraw their submissions at any time.However, if the same paper is also submitted forother competitions and wins other award(s) be-fore the ICSA announces the winners for the sym-posium, the author should notify the committeetimely and withdraw the paper.
Review and Selection Process Members of theStudent Award Committee will receive blindmanuscripts from the Committee Chair and reviewthem based on the following criteria:
• The manuscript should be well motivated byan application to the specific field(s);
• The methodology developed should be appli-cable to the motivating problem. Inclusion ofan application to a practical study will be fa-vorably considered;
• Organization and clarity of the presentationwill be considered as well.
Awards Up to eight student award winners (fiveStudent Travel Awards, one Jiann-Ping Hsu Phar-maceutical and Regulatory Sciences Student PaperAward, and two ASA biopharmaceutical Awards)will be selected. Each winner will receive a plaque,an award for travel and registration reimbursementup to $1000 or a cash award of $550, whicheveris bigger, as well as a free registration for a shortcourse. Winners will be notified around April 30,2014.
Student Award Committee Chair Wen-qing He, University of Western Ontario,[email protected].
© www.icsa.org 35
2014 ICSA Applied Statistics Symposium January 2014 Vol.26/1
Shortcourses at 2014 ICSA/KISS JointSymposium
There will be 3 one-day shortcourses and 4 half-day shortcourses at the 2014 ICSA/KISS joint sym-posium:
1. Recent Advances in Bayesian Adaptive Clin-ical Trial Design (Peter F. Thall & Brian P.Hobbs); one day.
2. Analysis of Life History Data with MultistateModels (Richard Cook & Jerry Lawless); oneday.
3. Propensity Score Methods in Medical Re-search for the Applied Statistician (RalphD'Agostino Jr.); one day.
4. ChIP-seq for transcription and epigeneticgene regulation (X. Shirley Liu); half day.
5. Data Monitoring Committees In Clinical Tri-als (Jay Herson); half day.
6. Analysis of Genetic Association Studies UsingSequencing Data and Related Topics (XihongLin & Seunggeun Lee); half day.
7. Analysis of biomarkers for prognosis and re-sponse prediction (Patrick J. Heagerty); halfday.
Recent Advances in BayesianAdaptive Clinical Trial Design
Presenters Peter F. Thall & Brian P. Hobbs, The Uni-versity of Texas M.D. Anderson Cancer Center.Course Length One dayOutline/Description This one-day short course willcover a variety of recently developed Bayesianmethods for the design and conduct of adaptiveclinical trials. Emphasis will be on practical appli-cation, with the course structured around a seriesof specific illustrative examples. Topics to be cov-ered will include: (1) using historical data in bothplanning and adaptive decision making during thetrial; (2) using elicited utilities or scores of differenttypes of multivariate patient outcomes to charac-terize complex treatment effects; (3) characterizingand calibrating prior effective sample size; (4) mon-itoring safety and futility; (5) eliciting and establish-ing priors; and (6) using computer simulation as adesign tool. These methods will be illustrated by ac-tual clinical trials, including cancer trials involvingchemotherapy for leukemia and colorectal cancer,
stem cell transplantation, and radiation therapy, aswell as trials in neurology and neonatology. The il-lustrations will include both early phase trials to op-timize dose, or dose and schedule, and randomizedcomparative phase III trials.References
• Braun TM, Thall PF, Nguyen H, de Lima M.Simultaneously optimizing dose and sched-ule of a new cytotoxic agent. Clinical Trials,4:113-124, 2007.
• Hobbs, BP, Carlin, BP, Mandrekar, S, Sargent,DJ. Hierarchical commensurate and powerprior models for adaptive incorporation ofhistorical information in clinical trials, Bio-metrics, 67: 1047–1056, 2011.
• Hobbs, BP, Sargent, DJ, Carlin, BP Commen-surate priors for incorporating historical in-formation in clinical trials using general andgeneralized linear models. Bayesian Analysis,7: 639–674, 2012.
• Hobbs, BP, Carlin, BP, Sargent, DJ. Adaptiveadjustment of the randomization ratio usinghistorical control data. Clinical Trials, 10:430-440, 2013.
• Morita S, Thall PF, Mueller P. Determiningthe effective sample size of a parametric prior.Biometrics. 64:595-602, 2008.
• Morita S, Thall PF, Mueller P. Evaluating theimpact of prior assumptions in Bayesian bio-statistics. Statistics in Biosciences. 2:1-17,2010.
• Thall PF. Bayesian models and decision algo-rithms for complex early phase clinical trials.Statistical Science. 25:227-244, 2010.
• Thall PF, Szabo A, Nguyen HQ, et al. Opti-mizing the concentration and bolus of a drugdelivered by continuous infusion. Biometrics.67:1638-1646, 2011.
• Thall PF, Nguyen HQ. Adaptive random-ization to improve utility-based dose-findingwith bivariate ordinal outcomes. J Biophar-maceutical Statistics. 22:785-801, 2012.
• Thall PF, Nguyen HQ, Braun TM, QazilbashM. Using joint utilities of the times to responseand toxicity to adaptively optimize schedule-dose regimes. Biometrics. In press.
About the Presenters Dr. Peter Thall has pioneeredthe use of Bayesian methods in medical research.
36 © www.icsa.org
January 2014 Vol.26/1 2014 ICSA Applied Statistics Symposium
He has published over 160 research papers andbook chapters in the statistical and medical litera-ture, including numerous papers providing inno-vative methods for the design, conduct and analy-sis of clinical trials. Over the course of his career,he had designed over 300 clinical trials. He has pre-sented 20 short courses and over 130 invited talks,and regularly provides statistical consultation forcorporations in the pharmaceutical industry. Hehas served as an associated editor for the journalsStatistics in Medicine, Journal of National CancerInstitute, and Biometrics, currently is an associateeditor for the journals Clinical Trials, Statistics inBiosciences, and is an American Statistical Associ-ation Media Expert.
Analysis of Life History Data withMultistate Models
Presenter Richard Cook and Jerry Lawless, Univer-sity of WaterlooCourse Length One dayOutline/Description Life history studies examinespecific outcomes and processes during peoples'lifetimes. For example, cohort studies of chronicdisease provide information on disease progres-sion, fixed and time-varying risk factors, and the ex-tent of heterogeneity in the population. Modellingand analysis of life history processes is often facil-itated by the use of multistate models. The aim ofthis workshop is to present models and methods formultistate analyses, and to indicate some currenttopics of research. Software for conducting analy-ses will be discussed and code for specific problemswill be given. A wide range of illustrations involv-ing chronic disease and other conditions will be pre-sented. Course notes will be distributed.
TOPICS: 1. Introduction; 2. Some Basic Quan-tities for Event History Modelling; 3. Some Illus-trative Analyses Involving Multistate Models; 4.Processes with Intermittent Observation; 5. Mod-elling Heterogeneity and Associations; 6. Depen-dent Censoring and Inspection; 7. Some Other Top-ics.About the Presenters Richard Cook is Professor ofStatistics at the University of Waterloo and holderof the Canada Research Chair in Statistical Meth-ods for Health Research. He has published exten-sively in the areas of statistical methodology, clin-ical trials, medicine and public health, includingmany articles on event history analysis, multistatemodels and the statistical analysis of life history
data. He collaborates with numerous researchersin medicine and public health and has consultedwidely with pharmaceutical companies on the de-sign and analysis of clinical trials.
Jerry Lawless is Distinguished Professor Emer-itus of Statistics at the University of Waterloo. Hehas published extensively on statistical models andmethods for survival and event history data, life his-tory processes and other topics, and is the author ofStatistical Models and Methods for Lifetime Data(2nd edition, Wiley, 2003). He has consulted andworked in many applied areas, including medicine,public health, manufacturing and reliability. Dr.Lawless was the holder of the GM-NSERC Indus-trial Research Chair in Quality and Productivityfrom 1994 to 2004.
Drs. Cook and Lawless have co-authored manypapers as well as the book "The Statistical Analy-sis of Recurrent Events" (Springer, 2007). They havealso given numerous workshops together.
Propensity Score Methods in Med-ical Research for the AppliedStatistician
Presenter Ralph D'Agostino Jr., Wake Forest Uni-versity School of MedicineCourse Length One DayOutline/Description The purpose of this shortcourse is to introduce propensity score methodol-ogy to applied statisticians. Currently propensityscore methods are being widely used in research,but often their use is not accompanied by an expla-nation on how they were used or whether they wereused appropriately. This course will teach the at-tendee the definition of the propensity score, showhow it is estimated, and present several applied ex-amples of its use. In addition, SAS code will be pre-sented to show how to estimate propensity scores,assess model success and perform final treatmenteffect estimation. Published medical journal arti-cles that have used propensity score methods willbe examined. Some attention will be given to theuse of propensity score methods for detecting safetysignals using post-marketing data. Upon comple-tion of this workshop, researchers should be ableto understand what a propensity score is, to knowhow to estimate it, to identify under what circum-stances they can be used, to know how to evaluatewhether a propensity score model“worked”, andto be able to critically review the medical literature
© www.icsa.org 37
2014 ICSA Applied Statistics Symposium January 2014 Vol.26/1
where propensity scores have been used to deter-mine whether they were used appropriately. In ad-dition, attendees will be shown statistical programsusing SAS software that will estimate propensityscores, assess the success of the propensity scoremodel and estimate treatment effects that take intoaccount propensity scores. Experience with SASprogramming would be useful for attendees.Textbook/References
• Rosenbaum P, Rubin DB. The central role ofthe propensity score in observational studiesfor causal effects. Biometrika. 1983; 70:41-55.
• D'Agostino RB Jr. Propensity score methodsfor bias reduction in the comparison of a treat-ment to a non-randomized control group. StatMed 1998; 17:2265-2281.
• Rubin DB. The design versus the analysis ofobservational studies for causal effects; par-allels with the design of randomized studies.Stat Med. 2007; 26:20-36.
• D'Agostino RB Jr., D'Agostino RB Sr. Estimat-ing treatment effects using observational data.JAMA 2007; 297(3) 314-316.
• Yue LQ. Statistical and regulatory issues withthe application of propensity score analysis tonon-randomized medical device clinical stud-ies. J Biopharm Stat. 2007; 17(1):1-13.
• D'Agostino RB Jr. Propensity scores incardiovascular research. Circulation 2007;115(17):2340-2343.
About the Presenter Dr. D'Agostino holds a Ph.D.in Mathematical Statistics from Harvard University.He is a Fellow of the American Statistical Associ-ation and a Professor of Biostatistical Sciences atthe Wake Forest School of Medicine (WFSM). Hehas been a principal investigator for several RO1grants/subcontracts funded by the NIH/CDC andhas served as the Statistical Associate Editor forArthroscopy (The Journal of Arthroscopy and Re-lated Surgery) since 2008 and has previously beenon the editorial boards for Current Controlled Tri-als in Cardiovascular Medicine, the Journal of Car-diac Failure and the American Journal of Epidemi-ology. He has published over 235 manuscripts andbook chapters in areas of statistical methodology (inparticular propensity score methods), cardiovascu-lar disease, diabetes, cancer and genetics. He hasextensive experience in the design and analysis ofclinical trials, observational studies and large scaleepidemiologic studies. He has been an author onseveral manuscripts that describe propensity scoremethodology as well as many applied manuscriptsthat use this methodology. In addition, during the
past twenty years, Dr. D'Agostino has made nu-merous presentations and has taught several shortcourses and workshops on propensity score meth-ods.
ChIP-seq for transcription and epi-genetic gene regulation
Presenter X. Shirley Liu, Harvard School of PublicHealth, Dana-Farber Cancer Institute, and Broad In-stitute.Course Length Half day.Outline/Description With next generation se-quencing, ChIP-seq has become a popular tech-nique to study transcriptional and epigenetic generegulation. The short course will introduce thetechnique of ChIP-seq and discuss the computa-tional and statistical issues in analyzing ChIP-seqdata. They includes the initial data QC, normaliz-ing biases, identifying transcription factor bindingsites and target genes, predicting additional tran-scription factor drivers in biological processes, inte-grating binding with transcriptome and epigenomeinformation. We will also emphasize the impor-tance of dynamic ChIP-seq, and introduce some ofthe tools and databases that are useful for ChIP-seqdata analysis.Textbook/References
• Park PJ. ChIP-seq: Advantages and challengesof a maturing technology. Nat Rev Genet.2009 Oct;10(10):669-80.
• Shin H, Liu T, Duan X, Zhang Y, Liu XS. Com-putational methodology for ChIP-seq analy-sis. Quantitative Biology. 2013.
About the Presenter Dr. X. Shirley Liu is Pro-fessor of Biostatistics and Computational Biologyat Harvard School of Public Health and Directorof the Center for Functional Cancer Epigenetics atthe Dana-Farber Cancer Institute. Her research fo-cuses on computational models of transcriptionaland epigenetic regulation by algorithm develop-ment and data integration for high throughputdata. She has developed a number of widelyused transcription factor motif finding (cited over1700 times) and ChIP-chip/seq analysis algorithms(over 8000 users), and has conducted pioneering re-search studies on gene regulation in development,metabolism, and cancers. Dr. Liu published over100 papers, including over 30 in Nature, Science orCell series, and she has an H-index of 50 according
38 © www.icsa.org
January 2014 Vol.26/1 2014 ICSA Applied Statistics Symposium
to Google Scholar statistics. She presented at over50 conferences and workshops, and gave researchseminars at over 70 academic and research institu-tions worldwide.
Data Monitoring Committees InClinical Trials
Presenter Jay Herson, Johns Hopkins BloombergSchool of Public Health.Course Length Half dayOutline/Description This workshop deals withbest practices for data monitoring committees(DMCs) in the pharmaceutical industry. The em-phasis is on safety monitoring because this consti-tutes 90% of the workload for pharmaceutical in-dustry DMCs. The speaker summarizes experienceover 24 years of working as statistical member or su-pervisor of statistical support for DMCs. He pro-vides insight into the behind-the-scenes workingsof DMCs which those working in industry or FDAmay find surprising. The introduction presents astratification of the industry into Big Pharma, Mid-dle Pharma and Infant Pharma which will be re-ferred to often in this workshop. Subsequent sec-tions deal with DMC formation, DMC meetingsand the process of serious adverse event (SAE) dataflow. The tutorial's section on clinical issues ex-plains the nature of MedDRA coding as well asissues in multinational trials. This will be fol-lowed by a statistical section which reviews and il-lustrates the various methods of statistical analy-sis of treatment-emergent adverse events, dealingwith multiplicity and, if time allows, likelihood andBayesian methods. The workshop's review of bi-ases and pitfalls describes reporting bias, analysisbias, granularity bias, competing risks, and recom-mendations to reduce bias. A description of DMCdecisions goes through various actions and ad hocanalyses the DMC can make when faced with anSAE issue and their limitations. The workshop con-cludes with emerging issues such as adaptive de-signs, causal inference, biomarkers, training DMCmembers, cost control, DMC audits, mergers andlicensing and the high tech future of clinical trials. .Text Herson, J. Data and Safety Monitoring Com-mittees in Clinical Trials, Chapman & Hall / CRC,2009.About the Presenter Jay Herson received his Ph.D.in Biostatistics from Johns Hopkins in 1971. Af-ter working on cancer clinical trials at MD Ander-son Hospital he formed Applied Logic Associates
(ALA) in Houston in 1983. ALA grew to be abiostatistical-data management CRO with 50 em-ployees when it was sold to Westat in 2001. Jayjoined the Adjunct Faculty in Biostatistics at JohnsHopkins in 2004. His interests are interim analy-sis in clinical trials, data monitoring committees,and statistical / regulatory issues. He chaired thefirst known data monitoring committee in the phar-maceutical industry in 1988. He is the author ofnumerous papers on statistical and clinical trialmethodology and, in 2009, authored the book Dataand Safety Monitoring Committees in Clinical Tri-als, published by Chapman Hall / CRC.
Analysis of Genetic AssociationStudies Using Sequencing Dataand Related Topics
Presenters Xihong Lin, Harvard School of PublicHealth, and Seunggeun Lee, University of Michi-gan.Course Length Half dayOutline/Description The short course is to discussthe current methodology in analyzing sequencingassociation studies for identifying genetic basis ofcommon complex diseases. The rapid advances innext generation sequencing technologies providesan exciting opportunity to gain a better understand-ing of biological processes and new approaches todisease prevention and treatment. During the pastfew years, an increasing number of large scale se-quencing association studies, such as exome-chiparrays, candidate gene sequencing, whole exomeand whole genome sequencing studies , have beenconducted, and preliminary analysis results havebecome rapidly available. These studies could po-tentially identify new genetic variants that play im-portant roles in understanding disease etiology ortreatment response. However, due to the massivenumber of variants and the rareness of many ofthese variants across the genome, sequencing costs,and the complexity of diseases, efficient methodsfor designing and analyzing sequencing studies re-main virtually important yet challenging.
This short course provides an overview of statis-tical methods for analysis of genome-wide sequenc-ing association studies and related topics. Top-ics include study designs for sequencing studies,data process pipelines, statistical methods for de-tecting rare variant effects, meta analysis, genes-environment interaction, population stratification,
© www.icsa.org 39
2014 ICSA Applied Statistics Symposium January 2014 Vol.26/1
mediation analysis for integrative analysis of ge-netic and genomic data . Data examples will be pro-vided and software will be discussed.Textbook/References Handout and references willbe provided.About the Presenters Xihong Lin is Professor ofBiostatistics and Coordinating Director of the Pro-gram of Quantitative Genomics at the School ofPublic Health of Harvard University. Dr. Lin's re-search interests lie in statistical genetics and 'omics,especially development and application of statisti-cal and computational methods for analysis of high-throughput genetic, and 'omics data in epidemio-logical and clinical studies; and in statistical meth-ods for analysis of correlated data such as longitu-dinal, clustered and family data. Dr Lin's specificareas of expertise include statistical methods forgenome-wide association studies and next genera-tion sequencing association studies, genes and en-vironment, mixed models, and nonparametric andseimparametric regression. She received the 2006Presidents' Award for the outstanding statisticianfrom the Committee of the Presidents of StatisticalSocieties (COPSS), and the 2002 Mortimer Spiegel-man Award for the outstanding biostatistician fromthe American Public Health Association. She is anelected fellow of the American Statistical Associ-ation, Institute of Mathematical Statistics, and In-ternational Statistical Institute. Dr. Lin was theChair of the Committee of the Presidents of the Sta-tistical Societies (COPSS) between 2010 and 2012.She is currently a member of the Committee of Ap-plied and Theoretical Statistics of the US NationalAcademy of Science. Dr. Lin is a recipient of theMERIT (Method to Extend Research in Time) fromthe National Institute of Health, which provides along-term research grant support. She is the PI ofthe T32 training grant on interdisciplinary trainingin statistical genetics and computational biology.She has served on numerous editorial boards of sta-tistical journals. She was the former CoordinatingEditor of Biometrics, and currently the co-editor ofStatistics in Biosciences, and the Associate Editorof Journal of the American Statistical Associationand American Journal of Human Genetics. She wasthe permanent member of the NIH study section ofBiostatistical Methods and Study Designs (BMRD),and has served on a large number of other studysections at NIH and NSF.
Seunggeun (Shawn) Lee is an assistant profes-sor of Biostatistics at the University of Michigan. Hereceived his Ph.D. in Biostatistics from the Univer-sity of North Carolina at Chapel Hill and completeda postdoctoral training at Harvard School of Pub-
lic Health. His research focuses on developing sta-tistical and computational methods for the analy-sis of the large-scale high-dimensional genetic andgenomic data, which is essential to better under-stand the genetic architecture of complex diseasesand traits. He is a recipient of the NIH Pathway toIndependence Award (K99/R00).
Analysis of Biomarkers for Prog-nosis and Response Prediction
Presenter Patrick J. Heagerty, University of Wash-ingtonCourse Length Half day.Outline/Description Longitudinal studies al-low investigators to correlate changes in time-dependent exposures or biomarkers with subse-quent health outcomes. The use of baseline or time-dependent markers to predict a subsequent changein clinical status such as transition to a diseasedstate require the formulation of appropriate clas-sification and prediction error concepts. Similarly,the evaluation of markers that could be used toguide treatment requires specification of operatingcharacteristics associated with use of the marker.
The first part of this course will introduce pre-dictive accuracy concepts that allow evaluation oftime-dependent sensitivity and specificity for prog-nosis of a subsequent event time. We will overviewoptions that are appropriate for both baseline mark-ers and for longitudinal markers. Methods will beillustrated using examples from HIV and cancer re-search and will highlight R packages that are cur-rently available. Time permitting, the second partof this course will introduce statistical methods thatcan characterize the performance of a biomarker to-ward accurately guiding treatment choice, and to-ward improving health outcomes when the markeris used to selectively target treatment. Exampleswill include use of imaging information to guidesurgical treatment, and use of genetic markers to se-lect subjects for treatment.Textbook/References
• Heagerty PJ, Lumley T, Pepe MS: Time depen-dent ROC curves for censored survival dataand a diagnostic marker. Biometrics 56:337-344, 2000.
• Heagerty PJ, Zheng Y: Survival model predic-tive accuracy and ROC curves. Biometrics.61(1): 92-105, 2005.
40 © www.icsa.org
January 2014 Vol.26/1 2014 ICSA Applied Statistics Symposium
• Saha P, Heagerty PJ: Time-dependent predic-tive accuracy in the presence of competingrisks. Biometrics, 66(4): 999-1011, 2010.
About the Presenter Patrick Heagerty is Profes-sor of Biostatistics, University of Washington, Seat-tle, WA. He has been the director of the center forbiomedical studies at the University of WashingtonSchool of Medicine and Public Health. He is oneof the leading experts on methods for longitudinal
studies including the evaluation of markers usedto predict future clinical events. He has made sig-nificant contributions to many areas of research in-cluding semi-parametric regression and estimatingequations, marginal models and random effects forlongitudinal data, dependence modeling for cate-gorical time series and hierarchical models for cate-gorical spatial data. He was an elected fellow of theAmerican Statistical Association and the Institute ofMathematical Statistics.
© www.icsa.org 41
Conversations January 2014 Vol.26/1
A Conversation with Dr. Robert O'NeillGordon Lan
IntroductionRobert O'Neill received his Ph.D. in Mathemat-
ical Statistics -- Biometry from Catholic Universityin 1971. Since then and to year 2011, he was em-ployed with the FDA and promoted from StatisticalReviewer to Team Leader and Director of Office ofBiostatistics. Under his leadership along with hisgreat efforts in promoting statistics in pharmaceu-tical and regulatory statistics, the size of the officeexpanded from less than a dozen to over one hun-dred and forty statisticians. In 2011, he assumeda newly created position of Senior Statistical Ad-visor in the Center for Drug Evaluation and Re-search, FDA. During his sabbatical leave in 1989,Dr. O'Neill visited the University of Basel, Switzer-land, where he delivered a lecture series on ``Topicsin Therapy Evaluation Review'' to over 400 partici-pants. The impact of the series on the pharmaceuti-cal industry stands clear. His leadership in pharma-ceutical statistics extended to an international plat-form when the pharmaceutical regulation moved tothe international harmonization stage. He led thedrafting of a sequence of ICH (International Com-munity Harmonization) guidance (ICH E9 and ICH
E11) with statistics relevance in clinical trials. In2013, Dr. O'Neill received the 2013 ICSA Distin-guished Achievement Award.
The interview presented below was organizedby Dr. Naitee Ting (Boehringer--Ingelheim) andDr. Yi Tsong (FDA) and was conducted by GordonLan (Johnson and Johnson) on April 29, 2013 at Dr.O'Neill's residence in Chestnut Hill, Massachusetts.
College and Joining FDAGordon: What led you to become a statistician andwhat led you to go to the FDA?Dr. O'Neill: I was always interested in mathemat-ics, so as an undergraduate, I had got my bache-lor's degree in mathematics from the College of theHoly Cross. It was a very theoretical program; Iwanted to continue in some form of math but I didnot want to go on into a pure theoretical area. Iwas not especially interested in the theorem--prooftype of approach --- for example some of the teach-ers that I had in college were into algebraic topol-ogy and group theory. So for graduate school Ilooked around the country for some applied pro-grams, not really knowing anything about the avail-able programs. I really didn't have any thoughtsabout the life sciences at that time. The person whowas advising me at Holy Cross said there was a pro-gram in Washington, DC at the Catholic Universityin biometry that I might be interested in that hadapplied and theoretical aspects to it, and the fellowwho was the head of that department was EugeneLukacs, a prominent person in the field of charac-teristic functions who had a book on characteristicfunctions. He has since passed away. Though thedepartment at Catholic University was theoreticallyoriented there were several applied professors instatistics, mathematical genetics and biological ap-plications. So I leaned toward the program in biom-etry and mathematical statistics as an applied areawith a strong theoretical basis. This progam waspart of the larger NIH sponsored program of fel-lowships in biometry that were very important totraining many of the statisticians of my age and co-hort.
So at Catholic University, I received a minor inbiometry and in mathematical genetics, where I hadto take biology courses, not medical courses but bi-ology courses and did fruit fly experiments to em-pirically demonstrate many of the Mendelian con-cepts critical to evaluating how dominant and re-
42 © www.icsa.org
January 2014 Vol.26/1 Conversations
Figure 1: Dr. O'Neill at his transfer party in 2011 with Drs. Bob Temple and Acting Director of Office ofClinical Pharmacology Shieu Mei Huang).
cessive genes are passed on to different generations.This was around the time that the whole DNAstructure was uncovered and that background ac-tually fit into some of the things I got involved inat FDA later on. So I went through the programand I happened to take a class in survey samplingfrom Chuck Anello , who was an adjunct professorat Catholic University and he was at that point intime in private industry, and had received his grad-uate degree from Johns Hopkins with Jerry Corn-field as his major professor.
We kept in touch and when I was graduating in1971 it was a tough job market and Chuck Anellohad gone to FDA about 1.5 to 2 years earlier, maybearound 1969, as FDA was changing, the law on reg-ulation of medicines had just changed and therewas a lot of reorganization going on at the Bureau ofDrugs which it was called at that time. Dr. Anellocalled me and asked me whether I would like tocome to work at the Bureau of Drugs, and I thoughtit was an interesting area where I could work withand get access to clinicians and chemists and otherscientists , so it was a multi-disciplinary environ-ment, and that's about all I knew about it. On theother hand, Steve Lagakos, a professor at Harvardwho I got to know early in my FDA career usedto say that I had told him that people had advisedme not to go to FDA in 1971 because there wereno interesting statistical problems at FDA. Lagakos
has since passed away, but I got familiar with Stevethrough the biometrics society and he asked me tobe a regional chairman on the regional advisoryboard, so I was active in the biometrics society atthat time because of Steve Lagakos.
So to recap, when I had finished up at CatholicUniversity with a degree in biometry and mathe-matical statistics, Chuck Anello, who had recentlygone to the Bureau of Drugs and taken over as direc-tor of the division of biometry at the time recruitedme. He had a very small staff and he was building aprogram with new Ph.D. recruits and he asked mewhether I wanted to come and I decided to go to theFDA figuring that it was probably an interesting op-portunity. There was really no established programyet -.it was a very interesting time because it wasonly 8 years or so from the new Harris--KefauverAmendments to the Food and Drug Act or law in1962 which made it necessary to demonstrate theefficacy of a drug in order to get on the market.The Harris--Kefauver amendments stated that youhad to demonstrate effectiveness of a new drug bymeans of substantial evidence based upon adequateand well-controlled trials. It was not until 1970 thata definition of adequate and well-controlled trialscame to be. This definition of substantial evidencefrom adequate and well-controlled trials was newto everyone and it took a while for FDA and the in-dustry to work through its implementation.
© www.icsa.org 43
Conversations January 2014 Vol.26/1
So I would venture to say that my reason for be-coming a statistician and then going to FDA andthen staying there, at least through the 1970s, wasn'tsomething that I planned out. One year at FDA ledto the next and I was given increasing levels of re-sponsibility. I started in the clinical trial side butthen around 1976--1977 I became a chief of a branchunit that supported all the areas outside the clin-ical trials, such as the safety area, which includeda lot of epidemiology, animal carcinogenicity, andgeneric drug bioequivalence work. So I learnedall that other part of the business because I was abranch chief responsible for those areas. I learneda lot of my epidemiology about case control studiesand observational study methods through that ex-perience. Eventually around 1979--1980, I becamedeputy director under Chuck Anello and took onmore responsible positions in the 1990's.
Career at FDAGordon: What did you enjoy most at the FDA andwhat you dislike?Dr. O'Neill: For the most part when I came to FDAin 1971, I learned everything I knew about clinicaltrials and the evaluation of clinical trials as well asall the analytic tools on the job. I learned how toanalyze data as well as how to evaluate a protocoland how to assess biases in the both the design andconduct of clinical trials. For my first 2 or 3 years ofexperience at FDA, I had to work real hard at learn-ing skills that I was not exposed to in school. I re-ally hadn't done a whole lot of data analysis in thegraduate program. nor had I used any computers.There was as a statistician at FDA who helped mea lot with how to approach the review of a clinicaltrial. His name was Charles Roberts who had hisPhD from either North Carolina or Georgia, as I re-call. He was a great mentor to me and he helpedme to deal with getting the clinical data in the trialsin electronic form which, in those days, we eithergot in paper form and needed to recode it or wouldget the data on computer tapes from companies. Icut my teeth by analyzing clinical trial data using aBMD program as it was the only program around atthe time to deal with complex designs such as un-balanced design, as multi-center trials were. Therewas no relevant SAS programs to speak of for thispurpose as I recall. BMD was a program developedat UCLA under sponsorship of NIH. There were anumber of programs in the BMD suite, one of whichwas called X64 which was a general linear modeland that was a workhorse. I used that a lot. That
was an interesting time because you had to do it allyourself by using the 64 column punch cards andrunning to the mainframe to get your analysis done.You would punch the cards up and then submitthem and hope that it ran because if it didn't run,you had to go run it again. That computer programcould handle the analysis of clinical trials of variousdesigns. It also handled covariates and analysis ofcovariance, which was important for the crossoverdesigns we analyzed, as that also was the beginningof the bioequivalence era and blood level studies tosupport generic drug approval.
When I recall that era, the analysis of most clin-ical trials required that you had to put the wholedesign matrix into the program code by hand; thatis write out the design matrix, write out all the treat-ment contrasts of interest, and type on punch cardsthe patient level data in a rectangular file structure.My first 3 years or so at FDA, from 1971 to 1974, Ireally had to learn a lot to be able to function effec-tively in that environment but that was a combina-tion of doing a lot of data analysis and understand-ing what were the problems with clinical trials.
Even though Chuck Anello was hiring morestaff, I think most of the reviewers like myself feltthat the program never had all the statistical re-sources it needed and that was a continual chal-lenge, if not a frustration. As I recall, Chuck's ini-tial buildup of the statistics program consisted ofabout 6--8 new PhDs that he had brought in. Soonafter I came, Bill Fairweather arrived and then SatyaDubey after that, then Ed Nevius and many otherswho contributed to the program. I think that manystatisticians who were there in the early 1970's leftfor other opportunities, probably by the end of the1970s or early 1980's. I was never bored with thework and never considered leaving during my firstyears. I was fortunate to be put into an environ-ment where I got exposed to a lot of different prob-lems early on; just not one area of clinical trials, asfor example epidemiological applications in safetyassessment and carcinogenicity studies, and bioe-quivalence.
One of the most important things I did, at least itseemed to me, in the early 1970s was to analyze a lotof phase III two period two treatment crossover de-signs which were what most of the industry wereusing at the time. There were not as many of theparallel group designs as today. And those two pe-riod two treatment crossover designs, though clin-ically appealing because the patient was used ashis/her own control, were virtually uninterpretablemost of the time because patients would not followthrough in the second period; they would drop out
44 © www.icsa.org
January 2014 Vol.26/1 Conversations
Figure 2: Dr. O'Neill explained some statistical issues to colleagues in FDA in a 2012 gathering.
for various reasons or not contribute full outcomesin the second period, so these designs became un-balanced because of missing outcomes data. Maybeyou could only analyze the first period if there wereadverse events, toxicity or safety issues occurring inthe first period that prevent useful information be-ing collected in the second period., and of coursethere were a lot of carryover effects observed, par-ticularly when placebo was administered second inthe sequence in which active drug was to be re-ceived first. During this period of time I was learn-ing a lot about that design and the statistics pro-gram was getting its feet on the ground in terms ofestablishing policies and procedures, and best prac-tices.
In the 1970's, the Bureau of Drugs had an ad-visory committee, called the Biometry and Epi-demiology Methodology Advisory Committee (BE-MAC), and that is when and how I got exposedto Marvin Zelen, Max Halperin, Jerry Cornfield,Sam Greenhouse, Seymour Geisser, John Gart, EdGehan, Byron Brown, and others, all the major bio-statistics players at that time. Chuck Anello wasthe first executive secretary of that committee and Ieventually became Executive Secretary around 1975or so. I took this role on in addition to my re-view work that I was responsible for. This role al-lowed me to interact with that committee and SamGreenhouse (Sam was a Chair, Jerry was a Chair).FDA statisticians presented issues of concern to us,or for which we needed advice, to that commit-tee. Around 1974 or 1975. we presented some con-cerns on such topics as dose response studies, onsequential analysis and repeated significance test-
ing in the early days of group sequential clinical tri-als, before you and others had done a lot of method-ological work in that area. In the 1970's there werefew if any texts at all for guidance, except for TedColton's Statistics in Medicine which we did finduseful in shaping our review of clinical trial reports.Ted had advised us on a few things with regardto reviewing the medical literature which was ap-plicable to New Drug Application (NDA) reviews.Chuck Anello was a classmate of Ted Colton's andTed would come down every once in a while andtalk about how you assessed the medical literatureand that influenced our thinking because at thattime we had no guidelines for the staff in terms ofhow to review NDA's or individual clinical trial re-ports. We were learning the process through mostof the 1970s and it wasn't until probably around1978 or so that we had developed internal guide-lines for our own staff to follow for how we wouldreview protocols and how we would review clinicaltrials. That was all developed as a result of the ex-perience that we all had collectively in the 1970's. Iwas able to contribute to guidance development be-cause I had come to FDA in 1971 and accumulatedsome experiences on process and content that couldbe incorporated into the early versions of the inter-nal guidances to FDA on protocol and study review.
But returning to the topic of the 1975 BEMACmeeting, building on my review experiences of theprevious three years, I made a presentation to theBEMAC committee about the two-period two treat-ment crossover design and its use in Phase III clin-ical trials for evidence of efficacy and safety of anew drug. We asked a series of questions to the BE-
© www.icsa.org 45
Conversations January 2014 Vol.26/1
MAC, and in their deliberations, a report was writ-ten to respond to those questions. They drafted areport which went back to the committee for delib-eration and Jerry Cornfield , who was chair at thattime, signed off on the report in 1977. Byron BrownMax Halperin and Sam Greenhouse all contributedto that particular report, but I wrote a fair amountof the draft in conjunction with Byron Brown. Soonafter that, Bill (Byron) Brown wrote an article inBiometrics that he published around 1980 essen-tially summarizing the work and expanding someof the methodology and conclusions. That publi-cation was really the first major paper on this issuefrom a regulatory perspective that was publishedafter Jim Grizzle's paper. Jim Grizzle was the chairof the department of biostatistics in North Carolina,and he had published an article that everyone usedto analyze the data from the crossover design. Buthis approach included a conditional test in whichone would test for carryover effect and if you passedthat test then you would go and combine the datafrom the first and second period. Subsequently a lotof people determined that approach could inflatethe Type 1 error but it really was not well recognizedat the time because the research hadn't been doneon that yet. Subsequent to the BEMAC crossover re-port and Bill Brown's article, there have been manyarticles and 3 or 4 books written on the crossover de-sign, especially motivated by the regulatory issue.But prior to the BEMAC report there was nothingavailable. So probably that was the first major con-tribution I made, and I considered it a uniquely reg-ulatory statistical problem , just as other FDA moti-vated problems would also become part of regula-tory statistical research efforts.
FDA's public advisory committee process be-gan around 1972 or so, and consisted of a publicpart, and a closed session part . Today all commit-tee meetings are public. It was an interesting partof the job for an FDA statistician to present at ad-visory committees, not just the BEMAC but othersubject matter committees such as the cardiopul-monary committee. As a statistician I was intriguedabout that environment because it was an interest-ing mix of having to know your statistics, analyzedata, write a report, and present your findings toadvisory committees in public. So, you had to beable to communicate to a multi-disciplinary clini-cal audience and that was a skill that helped mefor the rest of my career --- namely, the ability tosummarize the key points and to be able to con-vince an external audience not just of statisticians. Ithink that experience helped me mature as a statis-tician, it helped me to understand how to mentor
other newer FDA statisticians about the job. We hada number of statisticians who didn't feel comfort-able in that environment; they usually came fromthe academic world, maybe researchers, and werenot as comfortable in interacting in a consulting en-vironment. As the FDA program was maturing andfeeling its way some people liked that environmentand some people didn't; some people thrived in it.For me, what became a very interesting part of thejob, as well as frustrating at times, was the mix ofall these responsibilities, including the interactionswith all the clinical people. This was often difficultin the 1970s because the clinical staff were not verysophisticated methodologically and they might notlisten to your advice. It could be frustrating consult-ing with clinicians who weren't sophisticated in thescience of clinical trials. Many clinicians in thosedays had come to FDA from their clinical practiceto the FDA to review data; they were not aware ofgood study design principles, they didn't have anappreciation for the clinical trial paradigm, and soa large part of the statistician's job was mentoringthem, training them. The job of a statistician wasnot only teaching the clinical people but also gain-ing their trustGordon: The FDA statistical units expanded a greatdeal in the 1970s and 1980s. Could you give us a re-view on what happened during that period of time?Dr. O'Neill: the decade of the 1970's there were 6medical review divisions at FDA; The Division ofBiometrics, as it was known consisted of 10 or 11statisticians and some data coders and some com-puter programmers. That environment supported6 medical review divisions. I was a reviewer forabout 3 years, then I took on a team leader role. Isupported the areas of cardiorenal and pulmonary,dental and surgical, and psychiatry and oncologyevery once in a while. I think one of the frustrat-ing aspects of the job was resources and the abilityto hire statisticians to help with the work. Duringthe 1970's the biostatistics programs was not ableto expand. As the review process matured and theteam approach to review matured, I was interestedin the balance of medical officers vs. statisticiansneeded to conduct a review but it was difficult toget folks to think that way. It was a continual chal-lenge to request and get new resources for the sta-tistical program. However, the bureau and sub-sequently the center management recognized thevalue of statistics over the years and started to sup-port the statistical program by giving FTEs to it sowhen crises arose, the statistical program receivedits share. That's the way the FDA usually got re-sources through crises or with new responsibility
46 © www.icsa.org
January 2014 Vol.26/1 Conversations
from the prescription drug user fee act, probablyover the course of a 20 year period and as the FDAgrew from 6--8 medical review divisions to 15, thestatistical program grew accordingly.
Currently in 2013, there are 15 medical reviewdivisions and an expanded post approval officeresponsible for evaluating adverse event reportssurveillance of safety, mostly with observationaldata , and an expanded generic drug review pro-gram as well as chemistry, manufacturing and like.The biostatistics program consists of about 170 staffin 7 biometric divisions to support these areas. Sothe growth of the biostatistics program paralleledthe growth of FDA itself as many new responsibili-ties were put on the FDA. This growth came aboutas a result increasing demands on the review pro-cess, new responsibilities placed on FDA, and var-ious public health crises. The interest in shorten-ing the drug review process stimulated the needfor more FDA resources and this was eventuallyaddressed by the Prescription Drug User Fee Act(PDUFA) that ushered in users fees paid by the in-dustry that then were used to hire more FDA staff.This PDUFA program has been renewed 5 timesand is now a little over 20 years in place. A con-siderable amount of statistical staff were recruitedto fulfill the PDUFA responsibilities.
For example, when the AIDS crisis occurredaround 1987, the anti-infective division split intoa separate anti-viral division to deal with work-load and new demands. It was interesting to ob-serve the different medical cultures in the differ-ent medical areas, which also impacted the type ofclinical trial designs used in those medical areas.Some medical areas did a lot of repeated measurestype of trials, other medical areas like anti-infectivedid more short-term acute studies and active con-trol trials. There were few large outcome trials ex-cept for the UGDP diabetes trial and the CoronaryDrug Project. Starting with the 1980's, FDA began toevaluate and gain experience with group sequentialtrials and got into the sequential analysis aspectsof trials. That's when data monitoring committeesstarted to take hold, and the industry started to do alot more large outcome trials. I recall a large publicmeeting that FDA initiated on the topic of admin-istrative looks in clinical trials, a practice that hadbeen occurring in the industry that FDA felt wasinappropriate and potentially contributed to oper-ational bias from unblinded access to interim studyresults. The industry wanted to monitor ongoingstudy data for administrative looks, with no allegedclaim to take any actions and, therefore, not hav-ing to pay a statistical price for the the interim un-
blinded analysis. Participating in that large pub-lic meeting were FDA statisticians and cliniciansand the Pharmaceutical Manufacturer's Association(PhRMA) (known as PMA at the time) statisticiansand clinicians, as well as some academics. We wereable to bring some major clarity to good practicesand principles in the area of administrative looksbecause the industry interest in this was primarilydriven by clinical and business concerns, such asgetting an read on whether to plan to build a man-ufacturing plant in Puerto Rico, if the trial lookedlike it would be a success and support an approval.Dave DeMets and others contributed to and rein-forced the idea that there was no free looks and thatadministrative looks requiring unblinded access torelative treatment differences should pay statisticalprice.Gordon: Other important contributions you havemade?Dr. O'Neill: Rather than consider that question lit-erally, I would like to answer it by discussing myperspective on how my job changed and maturedand how that impacted my FDA career. Early on,I felt that it was very important for FDA statisti-cians to get outside the four walls of FDA to tell peo-ple what was going on in the regulatory process sothat externally there was a better understanding ofFDA thinking and policies. As a result of that per-spective, I accepted, in conjunction with FDA ap-proval, external invitations and attended many pro-fessional meetings to speak on FDA related statisti-cal topics. One of the first meetings I attended witha pharmaceutical statistics focus was the PrincetonConference, which is now known as the DemingConference held in Atlantic City. I gave severaltalks there in the mid 1970's. I felt that in additionto the FDA reviewer role, the other part of the job,or dimension of the job that was important to be-ing effective was to have an outside exposure to thestatistical world. This as I have said, is because fewpeople understand the regulatory statistics model.FDA statisticians usually interacted with the indus-try sponsors and statisticians who came to FDA butusually those discussions were not made public. Ittook a long time, probably through the mid 1980sbefore there were open public professional meet-ings about the biostatistical pharmaceutical devel-opment problems. Now we have about 10--15 dif-ferent journals that have a lot of biopharmaceuticalstatistical content but in the 1970s and 1980s therewas less public discussion.
The other major aspect of my career dealt withmanaging a statistical program, helping it to growand guiding its direction within the FDA environ-
© www.icsa.org 47
Conversations January 2014 Vol.26/1
ment, and mentoring new division directors, andinteracting with the clinical and scientific leaders inCenter for Drug Evaluation and Research (CDER).When I first became the Director of the Division ofBiometrics, there was only one division with sev-eral units known as branches. With the growth ofthe program, other statisticians had to be groomedas leaders and statistical managers. The single divi-sion grew to an office of seven Divisions of Biomet-rics with about 170 people. One of the most chal-lenging aspects of the scaling up of a statistical pro-gram from a single division to an office was assur-ing the quality of the statistical reviews and adviceand assuring the consistency of the statistical adviceacross seven divisions. Putting policies in place andhaving a statistical team of senior statisticians to as-sure this quality and consistency was the challengeof an expanding statistical program.
Recruiting statisticians to FDA and to CDERin particular has always been challenging espe-cially when FDA was competing for new Ph.D. levelstatisticians with the NIH and academia, and withthe pharmaceutical industry which substantially in-creased its statistical community in the last 20 to 25years. Because a statistician considering a positionat FDA might not wish to be a full time researcheror a full time reviewer, the other challenge was tocreate an environment at FDA that allowed for re-view and research. That's what made it appealingfor statisticians who wanted to be a reviewer andalso to publish and wanted to contribute in the re-search arena to make it interesting for them.
I felt that it was important that the science ingeneral and the statistics in particular conducted byFDA be respected and that the scientific staff getrecognition for what they were doing and to allowthem to develop their review side as well as theirresearch side. In many ways , I think that goalwas reached over the years. We have had a lot ofvery good people come through the program go-ing back to the 1970s, 1980s and 1990s, probablystarting with Gordon Pledger, David Hall, and VernChinchili who left to teach at Penn State.
So bringing into FDA people who could both re-view and contribute to the literature in applied sta-tistical theory was important and I think that is an-other value to the statistical program. In my mind,regulatory research and regulatory statistics in par-ticular, is a uniquely different area that has yet tobe fully recognized in the academic world. So oneof the things I am trying to do now in my new rolehere is to promote courses and maybe a program inregulatory statistics.
In my evolution at FDA, I guess you could say
I was a statistician, a reviewer, a mentor of otherpeople, a manager, a spokesman and a marketer forthe program. I felt it was just as important to playthat role inside of FDA as it was to play that roleoutside of FDA. I accepted speaking requests andtried not to turn them down unless it adversely im-pacted my FDA responsibilities. I felt it was impor-tant to be seen and heard and help contribute in anarea. That was always the appeal of the job and, Ithink, the appeal of the FDA to many statisticiansbecause that is the reason statisticians stay at FDAin my view. That's not to say there aren't frustrat-ing times. Most FDA staff recognize that you haveto be careful about distinguishing your own opin-ion from something that might be perceived as FDApositions.
International ImpactsGordon: You spent an extended length of timewith the European and Japanese regulatory agen-cies. Could you summarize these experience andthe impact of the changes made in these regions?Dr. O'Neill: Chuck Anello, as Director of the statis-tical program, had spent a sabbatical year abroad inLondon in the end of the late 1970's or early 1980's.He worked with Martin Vessey who was a promi-nent researcher and epidemiologist. Chuck wasvery interested in epidemiology and the safety oforal contraceptives, so he went to Oxford to workwith Richard Peto and Martin Vessey. He enjoyedthat experience a lot and it influenced my thinkingabout how FDA could continue to interact with andimpact European statisticians dealing with pharma-ceutical regulatory statistical issues. There was aprogram at the Health and Human Services that al-lowed me to do something similar to what ChuckAnello had done. Sometime in the middle 1980s Istarted to think that it would be interesting to go toEurope to learn about their processes and to per-haps influence the role of statisticians in the regula-tory process.
I applied for a year in Europe as a work studyeffort. The rationale supporting my applicationcame from what I considered to be a great mis-understanding resulting from second-hand infor-mation that was filtered to the European pharma-ceutical industry by their U.S. colleagues. I feltthis situation adversely impacted how the indus-try perceived FDA's statistical policies and prac-tices. This filtered information was detrimental toharmonization and to the statistical profession, inmy mind. I also came to realize that there were no
48 © www.icsa.org
January 2014 Vol.26/1 Conversations
full time statisticians employed in any of the regu-latory agencies in Europe at that time in the 1980s,which was amazing to me having experienced whata statistician in drug regulation had to face in theUnited States. There were no statisticians in reg-ulatory agencies in London, nor in Germany, norFrance, the three major players in the European sys-tem. I thought this was very unfortunate and detri-mental to the statistical profession as it pertains toregulatory statistics. I thought that at least the U.Kwith its history of experience and development ofmedical statistics would be resourced, but that wasnot the case. In the U.K. there was an external advi-sory statistician by the name of David Jones, who Igot to know pretty well. David was a statistician atSt. Georges Hospital and he supplied statistical ad-vice to the U.K regulatory body when they neededstatistical input on license application reviews. Hewas called upon as an advisory committee memberbut did not have full time statisticians within the UKregulatory agency to confer with.
Europe was moving towards EC1992 in whichthe European community wanted to have freemovement of goods, people and services, and thisincluded development of a European wide cen-tral regulatory process for some centrally evaluateddrugs. I knew that some officials were writing guid-ances relevant to clinical trial design and evaluationin the regulatory setting for evidence, but the guid-ances that I was aware of were being written by clin-ical pharmacologists and not by clinical statisticiansso they lacked considerable substance and statisti-cal sophistication.
Recognizing this and recognizing that some-thing needed to be done, my proposal for the workstudy program was to spend a year at the Depart-ment of Research in the medical school at the Uni-versity of Basel in Switzerland. Dr. Fritz Buehler, acardiologist and chair of the department offered mean office and helped me shape my plans. My ideawas to do a series of seminars which would provideme access to key players in Basel as well as otherplayers in Europe. This resulted in a lecture seriescalled Topics in Therapy Evaluation Review. I gavelectures on Thursday afternoon. I thought theremight be 30-40 people signing up for the lectureseries but surprisingly about.400 people signed upfor it. The lectures occurred every other Thursdayfrom around 4--6:30. I did the first several of lec-tures then we would open it up for floor discussion.Then I started to invite other scientists and statis-ticians from Germany, Paris and London to give alecture or be on a panel. I invited John Lewis fromLondon and a German regulator from Berlin. When
I returned to the United States in 1990, Fritz Buhlerdecided the lecture series should be expanded sothe European Course in Pharmaceutical Medicine(ECPM) was born. The ECPM is a degree grantingprogram which over the last 20 years has graduatedover 2000 scientists all of whom have contributed ina positive way to improving drug development anda better understanding of the skills needed in drugdevelopment as well as in regulation. I still lecturefor one week every 2 years in one of the modules.
I am most proud of the fact that the lecture seriesbecame a degree-granting course which influencedthe next 25 years post-grads in Europe who areprimarily not statisticians but other disciplines. Inaddition, this experience in Basel impacted my in-volvement in the International Conference on Har-monization (ICH) which began in the next year af-ter I left Basel. Also I felt my Basel experience al-lowed me to influence the hiring of statisticians insome of the European regulatory agencies. For ex-ample, I was invited to give a mini version of lec-tures in London, Berlin and Paris. When in Londonat the London School of Hygiene, I attended a meet-ing with many prominent statisticians from the U.K. about how to impact the hiring of statisticians inthe British regulatory agency. A letter in the Journalof the Royal Statistical Society helped to have JohnLewis recruited and he was allowed to hire a groupof 5--6 people.
A similar thing happened in Germany whenJoachin Roehmel was recruited and he created a bio-statistics program in Berlin. The impact of those re-cruitments of senior statisticians in Europe was sub-stantial because they were responsible for the initialdrafts of statistical guidance that ultimately becameICHE9. That international guidance could not havebeen written had there not been regulatory statis-ticians in place during the 1990's, as ICHE9 wassigned off in 1998 as one of the last clinical trial ef-ficacy guidances developed.
With respect to my involvement with Japan andJapanese statisticians, that came about as a resultof the ICH process. I was topic leader for twoICH guidances, one of which is the E5 guidanceon the role of ethnic factors in the acceptance offoreign clinical data. That experience involved mewith many Japanese statisticians and their regula-tory agency so I had the opportunity to visit Japanmany times and to impact their understanding ofthe role of statisticians in regulatory review.
In addition to my ICH experience, our FDAbiostatistics program benefited from two Japanesestatisticians who spent time at FDA. Dr. MasaTakeuchi, who had spent 7 years with us at FDA,
© www.icsa.org 49
Conversations January 2014 Vol.26/1
returned to Japan to help improve the role of statis-ticians in their regulatory process. He is currentlyat Kitasato University educating the next generationof Japanese biostatisticians. Currently, as you knowwithin the ICSA, these experiences are now occur-ring within the Asian Pacific Economic Community(APEC) region.
So to recap this foreign study experience, I wasfortunate that I was given the opportunity to spendthe year in Europe, which eventually influenced myinvolvement with ICH and with Japan, hopefullyimproving the role of statisticians in the regulatoryprocess.
ICSA Involvement
Gordon: Tell us about your ICSA keynote speeches.Dr. O'Neill: I tried to identify some of the talks Ipresented to the ICSA. I believe the first one I gavewas in 2002 on the topic of Strategies for ManagingRisk Associated with New Drugs: A Statistical Per-spective on the Issues. That, I believe, was a keynotetalk. Then I gave a talk on Scenario Planning as aTool to Increase the Success Rate of Phase III trialsand to Enhance Drug Development in 2005 at theICSA meeting. In your 2008 meeting I gave a talkon Preparing for a Career in Regulatory Biostatis-tics --- What does it take? And there was a publi-cation in your ICSA journal on the Challenges forBiostatistics in Medical Product Development, In-novation and life regulation. So I have been fairlyactive in a number of the meetings and activities ofICSA.
As you know, the ICSA is an extremely activesociety that has become pretty strong in a relativelyshort period of time when you think about whenthe ICSA was created in 1989.
(Gordon explained the ICSA early times andhow the conferences came to be.)Dr. O'Neill: I think a lot of it had to do with thenumber of Chinese statisticians that we were ableto recruit to the FDA and as they gained more ex-perience at FDA and remained with FDA even now,ICSA benefited from those FDA statisticians attend-ing ICSA meetings, participating, organizing ses-sions and taking on leadership roles in the ICSA. Soin some ways there was a huge regulatory presencein the ICSA just by the membership. As you know,people like Yi Tsong and Sue-Jane Wang, Jim Hungand Greg Soon and others that I have not mentionedhave been very active in the ICSA, as have beenstatisticians in other FDA centers.
Advices for Young Statisticians
Gordon: Which parts of your personality and skillmake you such a big success in medical researchand drug development?Dr. O'Neill: Well, I never considered myself amethodologist, but I've written a lot with a reg-ulatory perspective that is just not to a statisticalaudience so I think my ability to work in a multi-disciplinary environment, particularly with clinicalpeople has helped. I also feel that having had aliberal arts undergraduate background at the Col-lege of the Holy Cross prepared me to operate inthe environment I chose to stay in. The liberal artsare not just about science but also about how tothink analytically about problems, and that is of-ten what you are faced with in regulatory settings.Early on, I tried to put myself in the position of thenon-statistician I was working with in order to un-derstand what it is, see why they have difficulty inunderstanding some of these complex issues. I havemade it a point to try to explain things as simplyas possible from a statistical perspective to the non-statistical audience. Also, I try to understand asmuch as I can about the clinical side of the thoughtprocess and logic --- really putting an effort intolearning as much about the problem as the clinicianknows, not about how to diagnosis or treat a dis-ease, but the clinical trial and pharmaceutical de-velopment area so as to provide meaningful solu-tions. I didn't always consider myself a good publicspeaker, presenter, or teacher but I worked on var-ious aspects of it over the years as you really haveto work hard at doing that. In addition to the partof the job that required a good command of statis-tical methods, analysis approaches, study designs,and philosophies of statistical inference, I enjoyedthe part of the job that forced me to clearly com-municate and be always understandable to others,I don't know whether I was that good at it earlyon. I had to develop that aspect of my skill mix soI think developing a consulting style is importanttoo, and trying to be forceful enough in what andhow you communicate your message at the righttime is equally important. That requires a certainlevel of confidence and comes with experience onthe job. You have to make sure that people are un-derstanding you; sometimes you can appear to be alittle more aggressive than you should be but thatside of the personality helped me to get the pointsacross. Being perceived as knowing what you knowand having conviction behind what you are saying,and using that to convince someone, spending timemaking sure that you say it right, and writing well
50 © www.icsa.org
January 2014 Vol.26/1 Conversations
--- that is important. I've probably written or co-authored about 90 papers which is not really partof the FDA job --- it was never the role of the job towrite research papers. I wouldn't say my major con-tributions have been research papers; they've beenpapers about relevant regulatory statistical issuesthat motivated me to write about them or to make acontribution to a complex issue that had regulatoryconsequences. Some things have been more heavilymethodology oriented, and many papers that havebeen methodologically oriented have been collab-orations with many smart FDA statisticians, all ofwhom have been able to provide motivations of theproblem.
It was important to expand my experience baseoutside of FDA, even though I also wanted toprovide an environment for the FDA statisticiansto do regulatory research. Chuck Anello, SatyaDubey, Bill Fairweather, Gordon Pledger, DavidZucker, Vern Chinchilli, and many other statisti-cians worked hard to promote a regulatory sta-tistical research opportunity for staff that supple-mented their regulatory review role, particularlywith encouraging external participation at meet-ings and professional societies.
For example, I got to know Peter Bauer very wellwhen I spent a year (1989--1990) in Basel, Switzer-land at the Department of Research in the medicalschool. I also had the opportunity to befriend manyGerman and European statisticians in that year,whose friendships have endured till now. When Pe-ter and I were at a meeting in Paris, I had an ideathat I had been thinking about for some time, andwas aware of a recent paper that he had publishedon combining p-values in the adaptive design fieldthat stimulated a solution to my idea. So I remem-ber sketching out on a napkin with Peter this ideaof the distribution of the p-value under the alter-native hypothesis and then we subsequently had adiscussion about power-spending functions as op-posed to alpha-spending functions, but the thingabout the p-value distribution, nobody had thoughtof it in that way. Most folks just considered thedistribution of the p-value as uniformly distributedover the interval zero one, but that's only underthe null. Under the alternative, the distribution isalso a function of the sample size and the partic-ular alternative and highly skewed. I said to Pe-ter nobody had written about this but I could usehis paper that he wrote with one of his colleaguesabout combining p-values in first and second stageof an adaptive design, and I told Jim Hung aboutthis and we started to write it up and work it up,and Jim was the lead author on that paper flesh-
ing out the ideas. I worked with Jim a lot on thattopic as well as other things but Jim was the intel-lectual engine on this just as Sue-Jane Wang and anumber of the other people that I've collaboratedwith. I have had the ideas but they sort of take itand run with it. And I've always enjoyed that envi-ronment at FDA --- it's a nice part of working withsome smart FDA people. It's an interesting part ofmy career to help shape that environment and wit-ness how other young statisticians coming to FDAcan seize the opportunities and create their own ap-plied research.Gordon: What advice do you give to young statisti-cian?Dr. O'Neill: Well, I would say work hard, nothingcomes easy, continue to learn, just because once youget your degree and get out of school, it is a contin-ual learning process. Don't sit back on your laurels.I think pushing yourself, trying to publish is im-portant. You don't have to publish every day but Ithink it's an important dimension. It's important toput yourself in positions where you have to stretchyourself a little, be willing to take some risks, don'tsit back in your comfort zone, seek the problem outand don't let the problem come to you. Alwaysbe inquisitive. The hard part is there are so manythings that you could do, figuring out which onesyou want to spend your time on because there's alot of interesting problems. I never had a plan towork at the FDA for 40 years. I didn't know where Iwould want to work. I didn't really know whether Iwanted to be a statistician. I sort of happened into it.It's like a number of statisticians of my age and co-hort who came to the field of biostatistics that way. Irecall having brief conversations on this topic withother very influential statisticians like Rich Simonand Dave DeMets about how they got into the fieldand stayed with it. You might say we just stumbledinto it. However, the times and the university sta-tistical programs are a lot different today and theymay better prepare you as well as provide more in-sight into what options are available after you grad-uate. Now there's a body of work out there in thisvery well defined field of biostatistics and regula-tory statistics but there are so many different as-pects to it. I would imagine you cannot cram mostof this into a single graduate school program, andcertainly not a course --- you just don't have enoughtime.
You just have to commit yourself to continuedlearning, and be willing to put the time in becauseyou will have to be good at it to make an impact.It's an after hours kind of thing. Read the journals,usually on your own time --- and there are so many
© www.icsa.org 51
Conversations January 2014 Vol.26/1
Figure 3: Dr. O'Neill played banjo.
of them now that it is hard to keep on top of new de-velopments and ideas. My wife used to think I wascrazy by reading the journals all the time at nightbut there is just so much material out there that atleast at that time it was interesting for me to readit, figure it out and see how it applied. I think thatas a young person there are many opportunities to-day in areas that are not just about clinical trials butmany other intersecting areas of epidemiology, per-sonalized medicine, risk assessment and the like. Ithink there is so much stuff going on in the moderndrug development world that there are new appliedfields opening up for statisticians who have the in-terest and desire. I think statistics in Epidemiologyand drug surveillance have a lot of opportunitiesfor people to do some new things there. I wouldsay that a modern statistician, even if he/she is inan academic role, needs to have a broad view ofthe world so that they can point their students indirections that they ought to go. One of the chal-lenging aspects to putting statisticians on our FDAadvisory committees is to ask the question aboutwhere the next generation of statisticians come fromwho can contribute in a public forum to decisionsthat are very important to public health. Usually inthese committees, there is one statistician sitting ona committee of 11 of 12 clinicians and the statistician
is trying to help their colleagues on the committeemake a decision based on data and help them sortthrough data and address difficult questions thatFDA needs advice on. The problem has been thatbecause of conflict of interest and things of this na-ture that FDA has to rely on academic statisticians tobecome advisory committee members. So we needacademic people who are not shy but rather will-ing to engage with their non statistical colleagues,as well as be aware of the regulatory standards ofevidence, guidances and other relevant material. Tome, you need to prepare yourself to take on thatkind of a role and I'm not so sure that we've doneas well as we could in preparing academic statisti-cians for this challenge. What I would like to be ableto do is to create a regulatory statistics and regu-latory science program to not only prepare peoplewho would be going into either an academic envi-ronment or a regulatory environment or a pharma-ceutical company environment but also to bring inthe whole component of how advisory committeemembers get trained, what they do, and how to pre-pare yourself for that kind of a role. We now havethe resources, technology, and opportunity with alot of recorded material and a lot of material thatcan be brought into a course such as case studies toillustrate how decisions are made, tough decisions,
52 © www.icsa.org
January 2014 Vol.26/1 Conversations
and things that aren't black and white and how tolook at different aspects of the problem.
Finding the mix of statisticians who are bothdisease and drug subject matter experts as wellas statisticians who know the approaches used inthose subject matter fields is important to our reg-ulatory process. What I mean by that are thereare many FDA committees needing knowledgeablestatisticians in different fields; for example, just indrugs alone there is a cardiorenal advisory com-mittee, anti-viral advisory committee, anti-infectiveadvisory committee, neuropharmacology advisorycommittee, dental and ophthalmic project, oncol-ogy, just to name a few. These committees deal witha whole spectrum of issues not only in drugs, but inmedical devices, vaccines, blood products, etc., andyou need to know the clinical medical subject mat-ter area and then bring your statistical thinking tothe issue to help committees deliberate. This pro-cess is conducted in a public forum and involves re-ceiving from FDA briefing packages that have a lotof information, a lot of data, and you have to figureout what is the crux of the problem. These are noton their face always black and white situations orthat easy to deal with. We've seen some statisticianswho were very good technically and were probablyfairly well-known but they might not necessarily bethat effective in a public forum and may be moreprone not to say anything than to say something.Gordon: Do you recall any difficult NDA that youhave seen?Dr. O'Neill: There are so many different exam-ples that could be an answer to this question thatit is hard to single out any one since all have theirfeatures. The most challenging are when thereis no precedent for a decision that depends upondata that may be subject to different interpretations.When the data is inadequate, that is easier to dealwith. Also, many issues are never clear cut and arenot resolvable without additional information thatmay take years to obtain. I can say that I was heav-ily involved in the Vioxx experience and that's anexample that played out over the course of aboutfour or five years. It had a lot of issues concerninginterpretation of clinical study designs, conflictingresults from multiple studies, meta-analysis of ran-domized trials of different designs, time dependentoutcomes and non-proportionality hazards, basiccalculations to estimate time dependent risks whichpeople still don't understand very well. The con-trol groups in the available studies were different,the disease studied were different, and the abilityto have a three year placebo control study in polypprevention was probably a serendipitous opportu-
nity to actually pin down a risk. I've used the Vioxxexperience as a poster child in a lot of talks I havegiven for many, many years, and I will probablycontinue to do so. The same can be said for thediabetes drug Avandia which was another difficultproblem to determine whether there were increasesin the risk of heart attacks and other safety issues.This problem leads to new guidance and criteria forpre and post approval criteria to assess the safetyof diabetes drugs. I've been involved in problemsfor which the industry sought dispute resolutionsfrom FDA when they did not agree with our deci-sions. I must say that the two period two treatmentcross-over problem which I spent a lot of time on inthe 1970s was difficult because you had to convincemany people that it was not a good idea to use thisparticular design for the purposes it was being usedand I think we were able to do that. Today, manyof the difficult issues aside from a drug decision,remain. As we all know, personalized medicine isforcing us to deal with drug response prediction,clinical study designs to best identify markers forresponse, and new approaches to dealing with in-ferentially sound subgroup analysis that is just nothypothesis exploration. In today's global develop-ment environment, a new challenge is the design,analysis and interpretation of the multi-regionalclinical trials which usually are large-outcome tri-als. When interpreting these trials, the heterogene-ity that may be observed presents challenges towhether the region specific results are generaliz-able. In some sense this is similar to whether theresult specifically should include or exclude a cer-tain subgroup or whether the whole result dependsupon this subgroup or whether a valid conclusioncan be reached. When there is only a large singlemultiregional trial the interpretation is difficult inthe absence of another confirmatory study, whichmay be unlikely to perform.
FDA recently approved a drug for which the re-sults were much better looking outside the UnitedStates than inside the United States and the criticalissue was whether the confounding of aspirin useand the dose of aspirin was a neutralizing factor inthat analysis and the decision on that inferentially.So statisticians have always had to address hetero-geneity and a lot of what is going on in global drugdevelopment now introduces various forms of het-erogeneity. We are likely to see it in more globalizedmedicine and personalized medicine --- that is verychallenging and difficult topic --- I don't think wehave a fix to it yet and we are going to have to figurethat out.
I think we are going to have to have a better un-
© www.icsa.org 53
Statisticians at Work January 2014 Vol.26/1
derstanding of prediction than what we do rightnow. I think people don't fully understand howhard prediction is from a perspective of personal-ized medicine. That is to say that you might beable to put individuals into categories where theymight have a higher probability of having a betterresponse to therapy but to think of it in terms of pre-dictive probability on a person level as opposed to apopulation level is challenging and we are not thereyet. How does that information get portrayed in la-beling of medical product and what is the evidenceneeded to do so, so that exploratory statements arenot oversold? How do we give people a better ideaof what their individual chances of a benefit are;I don't think we have the framework for that yet.We are going to have to deal with that in person-alized medicine, and I think of all the challengesin adaptive designs in trying to bring Bayesian andFrequentists thinking together. It still remains achallenge and I think we cannot have two statisticalcamps that are thinking about different aspects ofthis because it is not just us --- the statisticians --- it'sthe non-statisticians who are saying what is this ar-gument that you guys are having? I think there hasto be a better marrying of the concept of control type1 error with also assessing the probability of the al-ternative hypothesis being correct which is both afrequentist and Bayesian concept. Characterizingtype 1 error control for compound null hypothesesand multiple hypotheses for which strong family-wise error control is needed, as faced in adaptivedesigns, is also in need of consensus for frequentistand Bayesian thinking.
Finally, while I feel it is important to be cautiousin a regulatory setting because the decisions have
such public health consequences, I also am con-cerned about a perspective that many do not want totake risks, for example in that they will not try theseadaptive designs if the FDA is perceived as to notthink they are acceptable as a source of evidence.There is this delicate balance between gaining expe-rience because without the experience you won't beable to have the comfort levels; it's not unlike the ex-perience in the early days of data monitoring com-mittee strategies, etc. It may have taken people 10--15 years to feel comfortable with that experiencebut if you didn't try it you didn't know. I think wehave interesting challenges right now that are dif-ficult. I don't think they are insurmountable. Andthe whole area of quantitative safety is an area thatis difficult because you have to balance between themore rigorous type 1 error control type of a prob-lem and the probability of missing it if there is arisk. You have multiplicity, you have a false discov-ery going on --- a lot's involved in it but I think thisis an area for young statisticians to develop a careerin.Gordon: Thank you for your time and patience. Itis a great honor for me to have this chance speakingwith you.
Gordon Lan, Ph.D.Senior DirectorQuantitative SciencesJanssen R&DJohnson & Johnson
Statisticians at the United NationsYongyi Min
Why work for the United Nations
It might sound like a cliché, but the reality isthat the main reason my colleagues and I chose
to work for the United Nations is the opportunityto help to create a better world. I would like tosee that my work has a direct connection to peo-ple's lives around the globe. Statisticians at theUnited Nations compile and disseminate global sta-tistical information for decision makers on formu-lating sound policies that make a real difference in
54 © www.icsa.org
January 2014 Vol.26/1 Statisticians at Work
the lives of millions of people. The diversity of peo-ple in the United Nations also makes it a very inter-esting working environment. I enjoy working withcolleagues from all backgrounds and cultures, whocan bring different perspectives, ideas, and expe-riences from their own countries and upbringings.The United Nations also provides an almost unpar-alleled opportunity to travel and work in differentareas of the world.
What statisticians do at the UnitedNations
Time flies. It seems like almost yesterday thatI started my first day in New York at the UnitedNations Statistics Division (UNSD). That was on afrigid January day back in 2004. Like many otherjunior statisticians, I started with participating inthe global data collection on environment statis-tics, which included questionnaire design, data val-idation, development and maintenance of databasesystems, data compilation, and data dissemination.Beside environmental data, UNSD collects and dis-seminates data on international trade, national ac-counts, energy, industry, and demographic and so-cial statistics gathered from national and interna-tional sources and provides a global data centre inthese areas.
Collecting and disseminating global statistics in-formation is only one of the four main functionsof UNSD. The second important function of the di-vision is the development and promotion of inter-national statistical standards. In each area, statisti-cians at the United Nations work together with ex-perts around the world to develop definitions, clas-sifications, and methods that can be used in thecollection and compilation of statistics at the na-tional level. For example, the Principles and Recom-mendations for Population and Housing Censuses pro-vides international guidelines for countries to con-duct censuses. The System of National Accounts 2008sets up the latest international statistical standardfor the national accounts.
The other two main functions of UNSD areproviding technical advice, training, and capacitybuilding programmes to member states in the de-velopment of national statistical systems and facil-itating the coordination of international statisticalprogrammes and activities. My current work onthe Millennium Development Goals (MDGs) mon-itoring reflects how important these two functionsare to the strengthening of global statistical systemsand the improvement of policy making for global
development priorities.In September 2000, the world leaders of all 189
United Nations Member States gathered togetherin New York and adopted the Millennium Decla-ration, a commitment by nations to build a bet-ter world with less poverty, hunger, and disease;a world in which mothers and children have agreater chance of surviving and of receiving aneducation, and where women and girls have thesame opportunities as men and boys. It promiseda healthier environment and greater cooperation--- a world in which developed and developingcountries work in partnership for the betterment ofall. The MDGs translate this commitment into aframework of eight goals and time-bound targets bywhich progress can be measured (http://www.un.org/millenniumgoals/).
Statisticians in UNSD, coordinating withstatisticians in 27 other international and re-gional organizations, are responsible for theglobal and regional monitoring of progress to-wards the MDGs. Together, we produce theannual MDG report ((http://mdgs.un.org/unsd/mdg/Resources/Static/Products/Progress2013/English2013.pdf) whichpresents an accounting of how the world is doingin meeting the MDG goals and targets. The MDGreport is the most viewed publication publishedby the United Nations. The group of statisticiansalso work on the improvement of data and report-ing systems for MDGs, discussing strategies to im-prove MDGs monitoring and enhance methods forthe assessment of progress, and setting prioritiesfor statistical capacity building on MDG monitor-ing.
My favorite part of my job is working directlywith countries to help them in improving their na-tional statistical systems. Every year we organizemany regional workshops to provide training to na-tional statisticians. In the past couple of years, I con-ducted regional workshops in Jordan and Bangkokon improving monitoring and reporting of MDG in-dicators for countries in the West Asia region, aswell as in the Asia/Pacific region.
Where statisticians work at theUnited Nations
The Statistics Division I work at belongs to theUnited Nations Secretariat at the UN Headquar-ters in New York, with around 130 staff members.If one is interested in working in a particular re-gion, the statistics division at one of the five regional
© www.icsa.org 55
Statisticians at Work January 2014 Vol.26/1
economic commissions of the Secretariat can be anideal place to work, be it Bangkok in the Asia andPacific region, or Santiago in the Latin American re-gion. If you are interested in finding a job or an in-ternship at the UN Secretariat, you can visithttps://careers.un.org/. The United Nations Fundsand Programmes need statisticians to collect dataand conduct analysis for particular areas, such asmeasuring the situation of children and womenin the United Nations Children's Fund (UNICEF),measuring educational situations in the United Na-tions Educational, Scientific and Cultural Organi-zation (UNESCO), and monitoring health relatedissues in the World Health Organization (WHO).In addition, The Funds and Programmes also haveopened regional and country offices all over the
world, and some of them require statisticians' ex-pertise and assistance.
The United Nations promotes and encouragesstaff mobility to develop its activities properly andto adapt its wide range of programmes and activ-ities to a constantly changing global environment.Therefore, statisticians have many opportunities towork in different locations on different areas.
Yongyi Min, Ph.D.StatisticianStatistics DivisionUnited Nations
Statisticians at MedImmuneNon-Clinical Biostatistics GroupBinbing Yu and Harry Yang
MedImmune, LLC, is the worldwide biolog-ics research and development arm of AstraZeneca,a global biopharmaceutical company with focuseson the discovery, development and commercial-ization of small molecule and biologic prescrip-tion medicines. MedImmune, headquartered inGaithersburg, Maryland, became a wholly ownedsubsidiary of AstraZeneca in 2007 and currentlyis the AstraZeneca's strategic science center with2500 employees in the US and UK. MedImmuneconducts innovative research and exploring novelpathways across several key therapeutic areas, in-cluding respiratory, inflammation and autoimmu-nity; cardiovascular and metabolic disease; oncol-ogy; neuroscience; and infection and vaccines.
Non-Clinical Biostatistics (NCB) group is lo-cated within the Department of Translational Sci-ences (TS) with specializations in experimental de-sign, statistical modeling, and prediction and sim-ulation. The diverse skills and experiences in ourgroup can be deployed for proper utilization of the``scientific methodology'' to aid in making soundinferences and decisions in drug research and de-velopment (R&D). Our group utilizes various sta-tistical software and programming tools as wellas some novel statistical methods to provide sup-port for drug discovery, development and reg-
ulatory issues. Many in the group have prioracademic backgrounds and have taught and con-ducted research in major research universities, havepublished in both statistical and multi-disciplinarypeer-reviewed journals, and stay current in latestdevelopments within the discipline as well as in theregulatory environment. The members of the NCBgroup are encouraged to participate in professionalstatistical societies and regularly organize sessionsand present in workshops and seminars attendedby statisticians from industry, government and reg-ulatory agencies, and academics.
Our specific mission is to encourage and pro-mote statistical thinking in research planning andstudy designs, function as in-house statistical ex-perts on statistical methodology and applications,and provide high quality statistical support to Med-Immune’s drug R&D projects. Our ultimate mis-sion is to contribute to maximization of R&D ef-ficiency within MedImmune so that the Companycan bring safe and effective drugs to needy patientsas quickly and efficiently as possible.
The NCB group was formally established in2003. NCB presently consists of 13 PhD-level statis-ticians supporting diverse research and develop-ment activities. Collectively we have a wide rangeof experience and expertise in clinical and non-clinical study design and analysis, biomarker iden-tification/disease phenotyping/patient segmenta-
56 © www.icsa.org
January 2014 Vol.26/1 Statisticians at Work
tion, clinical, diagnostic and analytical assay de-velopment, formulation and process development,quality control and manufacturing, regulatory fil-ing, post-marketing safety analysis, predictionmodeling and simulation and statistical research.
We offer optimal solutions to complex and chal-lenging R&D problems through statistical consult-ing, data mining, modeling and simulation, soft-ware development, statistical training and collab-orative research. Specifically, we provide statisti-cal support in areas of drug discovery, pre-clinicaldevelopment, biomarker identification, analyticalmethod and process development, manufacturing,quality control, and regulatory submissions. We ac-tively collaborate and assist in study design, dataanalysis, interpreting results, data mining to extractkey knowledge, computational tool development,and scientific publications. We initiate and conductcollaborative and innovative statistical research topromote scientific excellence.
As statisticians in the pharmaceutical industry,we have the good fortune and opportunities to workwith individuals from different scientific and reg-ulatory disciplines on complex issues that do nothave easy solutions. We also often operate un-der compressed timelines. The diverse and time-sensitive nature of daily problems requires us tohave not only a solid grounding in both statisticsand sciences, but also a high level of problem-solving ability. At its heart statistical competencyis how to quickly parse a deluge of information, fil-ter out noise, and see clear signals. Furthermore,interpersonal skills are must-have qualities to workeffectively in collaborative team environments.
Statisticians bring tremendous value to everyaspect of drug R&D in the pharmaceutical indus-try, which is strictly regulated and faces unprece-dented challenges. Phenomenal amounts of com-plex data are generated daily. How to design ex-periments to maximize information at a low cost,how to parse the seemingly messy data to facilitatesolid decision-making, and how to control randomfluctuation so as to ensure compliance to obtain reg-ulatory approval is the main role of a statistician.Here is one typical example that we have workedwith our colleagues. The project's objective was toensure regulatory compliance in bioassay testing.Potency of a biological therapeutic is often deter-mined relative to a reference standard, such as via
the ratio of EC50 values. (EC50 is half maximaleffective concentration, referring to the concentra-tion of a drug, antibody or toxicant which inducesa response halfway between the baseline and max-imum after a specified exposure time.) Measure-ment of relative potency is only meaningful if thetest sample behaves as a dilution or a concentrationof the reference standard and exhibits a parallel re-lationship to the reference standard. Such similarityis called parallelism. Both United States Pharma-copeia (USP) and European Pharmacopeia requirethe evaluation of parallelism, and a new statisti-cal method called equivalence test is recommendedin recently revised USP Chapters <1032> Designand Development of Biological Assays and <1034>Analysis of Biological Assays. However, imple-mentation of the method can be a nightmare for lab-oratories that do not have statistical and softwareengineering expertise. We conducted both method-ological research and software development overa span of three years. By bringing together dif-ferent skills and expertise, our team brought freshinsights into the challenging issue. The joint ef-fort led to several publications and, most impor-tantly, a customized assay analysis template basedon a fully GMP-compliant software package. (GMP,Good Manufacturing Practices, refers to guidelineslaid down by agencies which control authorizationand licensing for manufacture and sale of food anddrug product or a medical device, active pharma-ceutical products and pharmaceutical excipients.)
Binbing Yu, Ph.D.Associate DirectorNon-Clinical BiostatisticsDepartment of Translational Sci-encesMedImmune, LLC.
Harry Yang, Ph.D.Senior DirectorNon-Clinical BiostatisticsDepartment of Translational Sci-encesMedImmune, LLC.
© www.icsa.org 57
Controversial Issues January 2014 Vol.26/1
Statistics for Big Data: Are StatisticiansReady for Big Data?John M. Jordan and Dennis K.J. Lin
Editorial: Dennis Lin gave a banquet speech atthe inaugural symposium of the ICSA CanadaChapter at Toronto (August 2-3, 2013), entitled“BIGstat@IC2SA.”It was so well received, I have in-
vited him to expand his talk for this article on BIGdata.I am pleased that he has agreed to do so and givenbelow is his view on Big Data. I sincerely hope thatyou will enjoy the reading as much as I did. HappyHolidays,
Ming-Hui Chen, 2013 President, ICSA
Abstract: After noting the relative absence of statisti-cians from the community of practice engaged with bigdata, we explain what big data is, how it's done, andwho's working with it. The paper then suggests thatstatisticians have much to contribute to both the intellec-tual vitality and the practical utility of big data. At thesame time, big data challenges statisticians to move out ofsome familiar habits to engage less structured problems,to become more comfortable with ambiguity, and to en-gage computer scientists in a more fruitful discussion ofwhat the various parties can bring to this new mode ofinvestigation.
Introduction
We seem to be living in a time of paradox. Asmore and more commercial firms and academicdisciplines proclaim their affinity for new kinds ofdata-driven modes of evidence and analysis, mem-bers of the professional statistical community fre-quently find themselves on the outside looking in.Even as the world discovers a set of tools, tech-niques, and attitudes lumped together in the phrase``big data,'' (see Figure 1) statistics as a discipline isall too often absent.
Figure 1: Relative interest as expressed by searchqueries at Google in the phrase "big data" (re-trieved 12/12/13 at http://www.google.com/trends/)
58 © www.icsa.org
January 2014 Vol.26/1 Controversial Issues
After explaining a bit about what big data is,how it's done, and who's working with it, we willsuggest that statisticians have much to contribute toboth the intellectual vitality and the practical utilityof this venture. At the same time, big data chal-lenges statisticians to move out of some familiarhabits to engage less structured problems, to be-come more comfortable with ambiguity, and to en-gage computer scientists in a more fruitful discus-sion of what the various parties can bring to thisnew mode of investigation. Theory and practice,history and innovation, and rigor and results allseem to be posed with increasing frequency as op-posites, when in the best investigation, they are infact held in creative tension.
What is Big Data?This might be a tricky question; for many in Sil-
icon Valley where this meme is taking shape, "BigData" is data that must be managed with a set oftechnologies called MapReduce. To an astrophysi-cist, or geneticist, or actuary, meanwhile, the no-tion might sound curious: these individuals seeenormous data sets addressed via other more tradi-tional means on a daily basis. There's no formal sizethreshold involved: data doesn't somehow become"big" at a petabyte, or whatever; the name insteadrefers more to a state of mind, in many cases to Web-scale user, network, and traffic data as encounteredand managed at Amazon, Facebook, Google, andYahoo. For statisticians, however, big data mightinitially be viewed as the class of problems involv-ing“large n and/or large p.”
Thus, big data has a nomenclature problem.Like so many other technologies -- smartphones,robots, or information security -- the popular namedoesn't really convey the essence of the situation.Yes, "big data" can involve very large volumes insome cases. But more generally, the phrase refers tonew kinds of data, generated, managed, and parsedin new ways, not merely bigger ones. While "BigData" is a vague phrase, there is some agreementthat it involves changes in scale along three dimen-sions (the three V's):
Volume: Whether it's your own hard disk space,the world's online video feeds, or a wealth of dig-ital sensors measuring many aspects of the planet,signs are abundant that data volumes are increas-ing steadily and substantially. The volume in bigdata can grow in part due to sensor traffic from the"Internet of Things," to social media, to more peo-
ple coming on line every day around the world,and from the increased use of geolocation, in partconnected to the rise of mobile communications.As a result, the size of today's datasets is growingenormously—from MB, GB, TB, all the way to ZB(zetabytes=1021 bytes). Accordingly, big data mustbe seen in some ways as a triumph of data stor-age, as the graph illustrates: hard disk storage isdropping in price faster than microprocessor per-formance (Moore's law) is improving. See, http://www.jcmit.com/diskprice.htm.
Variety: Big data is not only a matter of biggerrelational databases. As opposed to the familiarnumbers related to customer ID, stock keeping unit(SKU), or price and quantity, we are living in anage of massive amounts of unstructured data: e-mails, Facebook "likes," Tweets, machine traffic, andvideo. Performing analysis of heterogeneous datatypes often strains both the information technolo-gies and the statistical toolkits involved.
Velocity: Overnight batch processes are gettingto be less and less tenable as the world becomesan "always-on" information environment. WhenFedEx can tell me where my package is, or Fidelitycan tell me my net worth, or Google Analytics cantell me my website performance right now, the pres-sure is on more and more other systems to do like-wise. In some instances that we will not cover herein depth, being able to analyze streams of data (in-cluding on stock exchanges), some of them very bigand very fast, has become an imperative.
These three V's can be displayed in the popular-ized graph below.
© www.icsa.org 59
Controversial Issues January 2014 Vol.26/1
There's an important point to be made up front:big data is not necessarily complete, or accurate, or true.Asking the right questions is in some cases learnedthrough experience, or made possible by better the-ory, or a matter of luck. But in many instances, bythe time investigators figure out what they shouldbe measuring in complex systems, it's too late to in-strument the "before" state to compare to the "af-ter." Signal and noise can be problematic categoriesas well: one person's noise can be a gold mine forsomeone else. Context is everything. Value is in theeye of the beholder, not the person crunching thenumbers. Thus it is temping to mention "value" as afourth V. However, this is rarely the case. Big data isbig, often because it is automatically collected. Thusin many cases, it may not contain much informationrelative to noise. This is sometimes called a DRIP—Data Rich, Information Poor—environment. In anyevent, the point here is that bigger does not neces-sarily mean better when it comes to data.
Accordingly, big data skills cannot be purely amatter of computer science, statistics, or other pro-cesses. Instead, the backstory behind the creationof any given data point, category, or artifact canbe critically important. While the same algorithmor statistical transformation might be indicated ina bioscience and a financial scenario, knowing themath is rarely sufficient. Having the industry back-ground to know where variance is "normal," for in-stance, comes only from a holistic understanding ofthe process under the microscope. We thus recom-mend an informal fourth V to be “Veracity” (oreven“Validity”): managing the randomization inbig data is indeed one great opportunity for statis-ticians.
Who is working on big data? Howare they doing it?
Techniques used in big data analysis
Analyzing large, diverse data sets requires newtools. It is important to note, however, that this
toolset can often be applied to data that might notqualify as "big." Note also that many of these tech-niques can be difficult to define with purity sincemany are only subtly different from adjacent tools.Furthermore, the tools are often used in combina-tion.
A/B testing. Similar to clinical trials, A/B testingis often used to compare website enhancements: asample population at a shopping site might get ared banner or a 10% off coupon while a second pop-ulation is shown a blue banner or a "free shipping"offer, then purchase results are compared across thetwo groups and successful enhancements are in-cluded in production systems.
Data fusion. Data fusion is “a process dealingwith the association, correlation, and combinationof data and information from single and multiplesources to achieve refined position and identity esti-mates, and complete and timely assessments of sit-uations and threats, and their significance. The pro-cess is characterized by continuous refinements ofits estimates and assessments, and the evaluationof the need for additional sources, or modificationof the process itself, to achieve improved results.”
Data Mining. Data mining is a broad umbrellaincluding a variety of techniques to detect patternsin large datasets, using different combinations ofstatistics, computer science, and database manage-ment. Some examples of data mining include thefollowing: (a) Association rule learning, whichseeks to discover interesting across variables. Thesetechniques generate possible rules, which mightseek to explain why certain products are bought to-gether within a process called market basket anal-ysis. (b) Classification attempts to place new datapoints into categories that have been determined byprevious analysis of a training data set. (c) Clus-ter analysis groups items into sets—it differs fromclassification in its lack of training data —in thiscase the properties of similarity are not known inadvance. (d) Regression is a classical statisticaltechnique that attempts to determine how the valueof the dependent variable changes when one ormore independent variables is modified.
Machine learning. This is a broad term that en-compasses many techniques with origins in com-puter science, particularly what has been called'artificial intelligence" for more than 50 years. Inthe most fundamental definition, machine learningseeks to build systems that can learn to recognizecomplex patterns, and then make decisions basedon these data patterns. (a) Natural language pro-cessing is one example of machine learning. Applesmartphones' Siri feature and the IBM Watson com-
60 © www.icsa.org
January 2014 Vol.26/1 Controversial Issues
puter that won at Jeopardy are two familiar exam-ples of NLP, which involves far more than speechrecognition. (b) Unsupervised learning seeks touncover hidden structure in unlabeled data. Clus-ter analysis is an example of unsupervised learn-ing. (c) Supervised learning uses training data to"teach" a system to find similar patterns in subse-quent data to which it is exposed. (d) Ensemblelearning is a specific type of supervised learningin which multiple predictive models are combinedto obtain better predictive performance than couldbe obtained from any one of the constituent mod-els. In this combinatorial aspect ensemble learningresembles the approach of data fusion. (e) Neuralnetworks are a class of computational models mod-eled on biological neural networks, most typicallythose within a brain. These systems are used to findpatterns in either supervised or unsupervised fash-ion, and work particularly well for finding nonlin-ear patterns. Neural networks have been success-fully used for fraud detection.
Network analysis. Network analysis has come toprominence in the past 20 years both in the pursuitof asymmetric warfare adversaries including terrornetworks, as well as in connection with digital so-cial networks such as Twitter or Facebook. Severaltechniques are used to map and describe relation-ships among discrete nodes in a network.
Optimization. Optimization takes known sys-tems, in a factory, financial institution, or hospital,for example, and seeks to analyze and correct lim-itations to speed, cost, or outcomes. Genetic algo-rithms (or more recently Particle Swarm Optimiza-tion, for example) can be used for optimization.
Geospatial Analysis. Many techniques are avail-able to analyze the spatial properties related to agiven data set. The easy availability of cell phonegeolocation data, for example, helps mobile net-work providers design cell tower placement. Thefree global GPS network means that latitude andlongitude data is readily available for most any-where on the planet.
Simulation (Computer Experiment). Numerouscomputer techniques allow the behavior of complexsystems to be modeled, informing everything fromforecasting to scenario planning. Monte Carlo sim-ulations are commonly used: they run thousands ofsimulations, each based on different assumptions,generating a histogram that represents a probabil-ity distribution of potential outcomes. On the otherhand, some computer experiments (expensive sim-ulations) could take a long time to run, special careis needed.
Time series analysis. Set of techniques from both
statistics and signal processing for analyzing se-quences of data points, representing values at suc-cessive times, to extract meaningful characteristicsfrom the data. Time series forecasting is the use of amodel to predict future values of a time series basedon known past values of the same or other series.Some of these techniques, e.g., structural modeling,decompose a series into trend, seasonal, and resid-ual components, which can be useful for identifyingcyclical patterns in the data. Examples of applica-tions include forecasting sales figures, or predictingthe number of people who will be diagnosed withan infectious disease.
Visualization. Spreadsheets are particularly inef-fective in conveying complex relationships withinlarge data sets. New techniques are emerging to cre-ate images, diagrams, or animations to both com-municate help people understand data. Interactiv-ity is becoming more common, allowing users of adata set to manipulate the analyses rather than be-ing confined to static, two-dimensional ink and pa-per.
Who is working on big data?To date, big data is primarily a joint venture
between IT professionals (who provide expertiseon database management, data cleansing, and net-working) and computer scientists, who are in thealgorithm business. In any given investigation(whether video rental prediction, public health, orcredit card or insurance fraud prevention), subject-matter experts also play a key role. The data cre-ation process, in particular, might require exper-tise in handheld devices, sensor networks, satelliteimaging, or other particular technical domains. Thebusiness process under discussion, whether voting,online dating, life insurance, or ball bearing produc-tion, also dictates the need for experts in the laws,physical constraints, and other unique attributes ofthe particular phenomenon.
For all of the reliance on such traditional statisti-cal strong points such as regression analysis, statis-ticians have not played leadership roles in the move-ment: when the U.S. National Science Foundationconvened a working group on the topic in 2012,zero statisticians were named to a committee of 100experts. Why do we see this incongruity betweensubstantial historical expertise and limited contem-porary relevance? Three possible reasons come tomind:
• Big data has compiled a track record address-ing poorly defined problems. Statisticians
© www.icsa.org 61
Controversial Issues January 2014 Vol.26/1
have grown accustomed to well structuredproblems, often using a single technique ina delimited domain. The practical problemssolved by Netflix, Amazon, and Google aremessier than most statisticians see on a reg-ular basis.
• The computational intensiveness of big datalies outside the expertise of most statisticians.Being able to understand the entire data lifecycle, from sensor design, through collection,ETL (extract, transform, load), and storage isone common shortcoming related to IT exper-tise. Knowing how to assemble then burnmassive numbers of compute cycles in a clus-ter or cloud scenario is a second, and largelyunrelated, set of skills outside the statistician'ssweet spot.
• Professional statisticians have found that ask-ing productive, if only partially solved, ques-tions does not generate research publications.In contrast, narrowly defined problems thatcan be solved used robust theoretical con-structs can be found in all the major journals.We statisticians always prefer to be preciselywrong (missing the forest but microanalyz-ing a tree) rather than approximately right. Incontrast, many big data projects seek to pre-dict user behavior (Amazon and Google aretextbook examples) and do not seek repeat-able scientific laws underneath a successfulassociation or prediction.
What kinds of statistics are neededfor big data?
Many statisticians commonly appear to believethat (i) big data is better than small data, and(ii) new methodologies are powerfully robust andwork well in most cases. Consequently, the so-called “the death of p-value” is claimed. How-ever, disasters kept happening. Is it because thatthe fundamental statistical thinking still applied, al-though the theories may not be straightforwardlyapplied? Big data cannot replace scientific/statisti-cal thinking. Data and information/knowledge arenot synonyms. Thus a wish-list for needed statisti-cal methodologies should have the following prop-erties.
• High-impact problems. Refining existingmethodologies is fine, but more efforts should
focus on working high-impact problems, es-pecially those problems from other disci-plines. Statisticians seem to keep missingopportunities: examples range from genet-ics to data mining. We believe that statisti-cians should seek out high-impact problems,instead of waiting for other disciplines to for-mulate the problems into statistical frames.This leads to the next item.
• Provide structure for poorly defined prob-lems. A skilled statistician is typically mostcomfortable and capable when dealing withwell-defined problems. Instead, statisti-cians should develop some methodologies forpoorly defined problems and help devise astrategy of attack. There are many oppor-tunities for statistical applications, but mostof them are not in the “standard”statisticsframe —it will take some intelligent personsto formulate these problems into statistics-friendly problems (then to be solved by statis-ticians). Statisticians can devote more effortsto be such intelligent persons.
• Develop new theories. Most fundamentalstatistical theories based upon iid (indepen-dently identically distributed) for one fixedpopulation (such as, central limit theorem, orlaw of large number) may need to be modi-fied to be appropriately applied to big dataworld. Many (non-statisticians) believe thatbig data leads to“the death of p-value.”Thelogic behind this is that when the sample sizen becomes really large, all p-values will besignificant —regardless how little the prac-tical significance is. This is indeed a goodexample of misunderstanding the fundamen-tals. Another good example is about “smalln and large p”where the sparsity propertyis assumed. First, when there are many ex-ploratory variables, some will be classified asactive variables (whether or not this is true!).Even worse, after the model is built (mainlybased on the sparsity property), the residu-als may be highly correlated with some re-maining variables —this contradicts the as-sumption for all fundamental theorems that“error is independent with all exploratory
variables.”New measurement is needed forindependence in this case.
Having those wishlist items in mind, what kindsof statistics are needed for Big data? For the reasonof casting a brick to attract jade (抛砖引玉), given
62 © www.icsa.org
January 2014 Vol.26/1 Controversial Issues
below are some very initial thoughts under consid-eration.
• Statistics and plots for (many) descriptivestatistics. If the conventional statistics are tobe used for big data, and it is very likely therewill be too many of them, what is the best wayto extract important information from thesestatistics? For example, how to summarizethousands of correlations? How about thou-sands of p-values? ANOVA? Regression mod-els? Histograms? etc. Advanced methodsto obtain “sufficient statistics”(whatever itmeans) from those many conventional statis-tics are needed.
• Low-dimension behavior. Whatever methodis feasible for big data (the main concern beingthe computational costs), its low-dimensionbehavior is always important to be kept inmind.
• Norm or Extreme. Depending on the prob-lem, we could be interested in either normor extreme, or both. Basic methods for bothfeature extraction (mainly for extremes) andpattern recognition (mainly for norm) areneeded.
• Methods for new types/structures of data. Asimple example would be “How to buildup a regression model, when both inputsand outputs are network variables?”Mostexisting statistical methodologies are limitedto numbers (univariate or multivariate), al-though there is some recent work for func-tional data or text data. There are more thatcan be done, if we are willing to open ourminds.
• Prediction vs estimation. One difference be-tween computer science and statistics meth-ods has to do with the general goal —whileCS people focus more on prediction, statisti-cians focus more on estimation (or statisticalinference). Take Artificial Neural Networks(ANN) as an example: the method can fit al-most anything, but what does it mean? ANNis thus popularly used in data mining, but hasreceived relatively low attention from statisti-cians. For big data, it is clear that prediction isprobably more feasible in most cases. Note: insome very fundamental cases, we believe thatstatistical inference remains important.
Big Data in the Real World
Skills
Here's a quiz: ask someone in the IT shop howmany of his of her colleagues are qualified to workin Hive, Pig, Cassandra, MongoDb, or Hadoop.These are some of the tools that are emerging fromthe front-runners in big data, web-scale companiesincluding Google (that needs to index the entire In-ternet), Facebook (manage a billion users), Amazon(construct and run the world's biggest online mer-chant), or Yahoo (figure out what social media isconveying at the macro scale). Outside this smallindustry, big data skills are rare.
Politics
Control over information is frequently thoughtto bring power within an organization. Big data,however, is heterogeneous, multi-faceted, and canbring performance metrics where they had not pre-viously operated. If a large retailer, hypotheticallyspeaking, traced its customers' purchase behaviorfirst to social media expressions and then to adver-tising channel, how will the various budget-holdersrespond? Uncertainty as to ad spend efficacy is asold as advertising, but tracing ad channels to pur-chase activity might bring light where perhaps it isnot wanted. Information sharing across organiza-tional boundaries ("how are you going to use thisdata?") can also be unpopular.
Technique
Given that relational databases have beenaround for about 35 years, a substantial body oftheory and practice make these environments pre-dictable. Big data, by contrast, is just being in-vented, but already there are some important dif-ferences between the two: Most enterprise data isgenerated by or about humans and organizations:SKUs are bought by people, bills are paid by peo-ple, health care is provided to people, and so on.At some level, many human activities can be under-stood at human scale. Big data, particularly socialmedia, can come from people too, but in more andmore cases, it comes from machines: server logs,point of sale scanner data, security sensors, GPStraces. Given that these new types of data don'treadily fit into relational structures and can get mas-sively large in terms of storage, it's nontrivial to fig-ure out what questions to ask of these data types.
© www.icsa.org 63
Controversial Issues January 2014 Vol.26/1
When data is loaded into relational systems, itmust fit predefined categories that ensure that whatgets put into a system makes sense when it is pulledout. This process implies that the system is de-fined at the outset for what the designers expectto be queried: the questions are known, more orless, before the data is entered in a highly structuredmanner. In big data practice, meanwhile, data isstored in as complete a form as possible, close toits original state. As little as possible is thrown outso queries can evolve and not be constrained by thepreconceptions of the system. Thus these systemscan look highly random to traditional database ex-perts. It's important to stress that big data will notreplace relational databases in most scenarios; it's amatter of now having more tools to choose from fora given task.
Traditional DatabasesTraditional databases are designed for a con-
crete scenario, then populated with examples (cus-tomers, products, facilities, or whatever), usuallyone per row: the questions and answers one can askare to some degree predetermined. Big data can beharvested in its original form and format, and thenanalyzed as the questions emerge. This open-endedflexibility can of course be both a blessing and acurse.
Traditional databases measured the world innumbers and letters that had to be predicted:zip codes were 5 or 10 digits, SKU formats werecompany-specific, or mortgage payments were ofpredictable amounts. Big data can accommodateFacebook "likes," instances of the "check engine"light illuminating, cellphone location mapping, andmany other types of information.
Traditional databases are limited by the com-puting horsepower available: to ask harder ques-tions often means buying more hardware. Big datatools can scale up much more gracefully and cost-effectively, so decision-makers must become accus-tomed to asking questions they could not contem-plate previously. To judge advertising effectiveness,one cable operator analyzed every channel-surfingclick of every remote across every household in itsterritory, for example: not long ago, such an inves-tigation would have been completely impractical.
CognitionWhat does it mean to think at large scales? How
do we learn to ask questions of the transmissionof every car on the road in a metropolitan area, of
the smartphone of every customer of a large retailchain, or of every overnight parcel in a massive dis-tribution center? How can more and more peoplelearn to think probabilistically rather than anecdo-tally?
The mantra that "correlation doesn't imply cau-sation" is widely chanted yet frequently ignored; ittakes logical reasoning beyond statistical relation-ships to test what's really going on. Unless thedata team can grasp the basic relationships of howa given business works, the potential for complexnumerical processing to generate false conclusionsis ever-present. Numbers do not speak for them-selves; it takes a human to tell stories, but as DanielKahneman and others have shown, our stories oftenembed mental traps. Spreadsheets remain ubiqui-tous in the modern enterprise but numbers at thescale of Google, Facebook, or Amazon must be con-veyed in other ways. Sonification -- turning num-bers into a range of audible tones -- and visualiza-tion show a lot of promise as alternative pathwaysto the brain, bypassing mere and non-intuitive nu-merals. In the meantime, the pioneers are both see-ing the trail ahead and taking some arrows in theback for their troubles. But the faster people, andespecially statisticians, begin to break the stereo-type that "big data is what we've always done, justwith more records or fields," the faster the break-through questions, insights, and solutions will re-define business practice.
PrivacyIt has been proven repeatedly that anonymous
data sets can be reverse-engineered, identifyingpeople who either did not know they were a partof a study or trusted the process. Elsewhere, thecheapness of computer data storage (as measuredby something called Kreider's law) combines withthe ubiquity of daily digital life to create massivedata stores recording people's preferences, medi-cations, travels, and social contacts. As big datatools continue to increase in power, and compu-tational capability increases, and algorithmic so-phistication increases, composing revealing data-driven portraits of tens of millions of people willbe possible, profitable, and troubling. Put access tothose portraits on wearable digital devices such asglasses, and the prospect of facial recognition by arandom stranger on the street is certain to becomean issue sooner rather than later. Big data prac-titioners may face calls for a professional code ofconduct much like the Hippocratic oath for doctors:first, do no harm. Statisticians can provide sorely
64 © www.icsa.org
January 2014 Vol.26/1 Controversial Issues
needed expertise in this discussion.
Summing upNo matter what the field of inquiry, the form
and scale of the data involved, the computationalinfrastructure, or the name attached to the project,statistics as a discipline has much to offer the bigdata effort. Recall the insight offered by Box,Hunter and Hunter 35 years ago: “Data have nomeaning in themselves; they are meaningful only inrelation to a conceptual model of the phenomenonbeing studied.”Today, there are those who seemto suggest that models are unnecessary, that givensufficient computing power, the relevant patternswill emerge, absent theory. Indeed, there are thosewho argue that hypothesis-driven scientific methodwill become outdated in a world in which p-valueno longer matters.
There is much to be gained from using a“scientific”approach, as opposed to an“algorithm”
approach, to big data—including data collection,data analysis, mode selection, and feature interpre-tation. Before pushing the button (to run an auto-matic algorithm), perhaps more scientific thinkingis needed beforehand.
Seeing big data in a wider historical perspectivemight be a useful way to end our discussion. Su-san Hockfield recently retired as president of MITafter having been the first life scientist to lead theinstitution. She has concluded that “The conver-gence of life sciences and engineering, I think, is go-ing to be the story of the 21st century, much as theconvergence of the physical sciences and engineer-ing was the story of the 20th century." 1 During theearly 1900s, “physicists decoded the fundamentalelements of the physical universe. They were es-sentially understanding the parts list of the physi-cal world —the structure of atoms, how electronstravel." Engineers discovered this‘parts list’andbegan to turn theory into practice: the micropro-cessor, laser, and wireless networking -- the build-ing blocks of the computer revolution -- are among
these engineers' legacies.By the 1950s, Hockfield continued, scientists in-
cluding Watson and Crick were decoding the struc-ture of DNA, and “the biological sciences beganto assemble a parts list for the biological universe.And engineers, in a very similar way, as they sawthe ‘parts list’ evolving, picked up those partsand incorporated them into applications.”This in-sight underlies her contention that the 21st centurywill be an era characterized by everyday implemen-tations of recently discovered conceptual buildingblocks. For our purposes, the question is clear: whoamong the many computational, statistical, IT, anddomain experts is doing either of these two tasks:describing the "parts list" of quantitative investi-gation and discovery, or doing the engineering toturn theory into everyday reality, accessible to themasses?
Let us conclude this article by a quote fromHahn and Hoerl:“This is a golden age for Statistics,but not necessary for statisticians.”We sincerelyhope that this article provides some personal viewsfor statisticians to be ready for the many changesthat are underway under the banner of big data.In the simplest form, our advice is to be more ag-gressive, helping or even leading other disciplinesto advance science and knowledge using tools thathave served us well for centuries, revised to meetthe needs of this new era.
John M. Jordan, PhD.Clinical professor of Supply Chain &Information SystemsThe Pennsylvania State [email protected]
Dennis K.J. Lin, Ph.D.University Distinguished Professorof Statistics and Supply Chain Man-agementThe Pennsylvania State [email protected]
1James F. Smith, "Spotlight: Susan Hockfield and the Magic of the Laboratory," Winter 2012-13 newsletter, Belfer Center forScience and International Affairs, Harvard Kennedy School http://belfercenter.ksg.harvard.edu/publication/22515/spotlight.html. Our thanks to Dr. Roger Hoerl for bringing Hockfield's statements to our attention and for his inspiring discus-sions.
© www.icsa.org 65
Upcoming Events January 2014 Vol.26/1
Upcoming EventsWorkshop on Dimension Reduction andHigh Dimensional InferenceJanuary 17–January 18, 2014Gainesville, FL, USAhttp://www.stat.ufl.edu/symposium/2014/index.html
Ordered Data Analysis, Models andhealth Research Methods: An Interna-tional Conference in Honor of H. N. Na-garaja for His 60th BirthdayMarch 7–March 9, 2014The University of Texas at Dallas, Richardson, TX,USAhttp://faculty.smu.edu/ngh/hnnconf.html
Seminar on Stochastic Processes 2014March 26–March 29, 2014La Jolla, CA, USAhttp://depts.washington.edu/ssproc/
The 3rd Workshop on Biostatistics andBioinformaticsMay 9–May 10, 2014Atlanta, GA, USAhttp://www2.gsu.edu/~matyiz/2014workshop/
The 6th International Statistics Forum atRenmin University of ChinaMay 24–May 25, 2014Renmin University of China, Beijing, Chinahttp://stat.ruc.edu.cn/news/tongzhi/2013/1231/674.html
11th International 69th Conference on Or-dered Statistical DataJune 2–June 6, 2014Mathematical Research and Conference Center inBedlewo, Polandhttp://bcc.impan.pl/14OrderStat/
2014 ICSA symposiumJune 15–June18, 2014Portland, OR, USAhttp://www.statkiss.org/icsakiss2014/
Third Joint Biostatistics SymposiumJune 27–June28, 2014Chengdu, Chinahttp://www.icsa.org/meetings/co-sponsorship/index.html
The third IMS Asia Pacific Rim MeetingsJune 30–July 3, 2014Taipei, Taiwanhttp://www.ims-aprm2014.tw/
The 2014 ICSA-China Statistics Confer-enceJuly 4–July 5, 2014East China Normal University, Shanghai, China
Building Statistical Methodology andTheory: An International Conference inHonor of Jeff C.F. Wu for His 65th Birth-dayJuly 7–July 9, 2014Huquan Hotel, Mile, Yunnan, Chinahttp://www.stat.purdue.edu/~sunz/Jeff_2014/index.html
2014 IMS Annual MeetingJuly 7–July 11, 2014Sydney, Australiahttp://www.ims-asc2014.com/
2015 ICSA symposiumJune 14–June17, 2015Fort Collins, CO, USA
66 © www.icsa.org
The 6th International Statistics Forum at Renmin University of ChinaMay 24-25, 2014, Beijing China
Scientific Committee
• Chen, Ming-Hui, University of Connecticut• Geng, Zhi, Peking University• Hu, Feifang, George Washington University, Renmin
University of China• Xu, Xianchun, National Bureau of Statistics of China• Yuan, Wei (chair), Renmin University of China• Zhou, Andrew, University of Washington, Renmin Uni-
versity of China
Organizing Committee
• Chen, Songxi, Peking University• Ji, Hong, Capital University of Economics and Business• Jin, Yongjin, Renmin University of China• Liu, Yang, Central University of Finance and Economics• Ma, Shuangge, Yale University• Meng, Shengwang, Renmin University of China• Pan, Fan, National Bureau of Statistics of China• Wang, Hansheng, Peking University• Wang, Xiaojun, Renmin University of China• Xian, Zude, National Bureau of Statistics of China, In-
stitute of Government Statistics• Yan, Jun, University of Connecticut• Zhang, Bo, Renmin University of China• Zhao, Yanyun (chair), Renmin University of China
Plenary Speakers
• Dey, Dipak, University of Connecticut• Huan, Jian, University of Iowa, Shanghai University of
Finance and Economics• Tian, Maozai, Renmin University of China• Wang, Hansheng, Peking University
Topics
• Economic Statistics• Social, Demographic, Environmental Statistics• Analysis of High Dimensional Data• Risk Management and Actuarial Science• Financial Statistics
• Biostatistics• High Frequency Data Analysis• Statistical Learning and Data Mining• Nonparametric Statistics• Bayesian Analyses• Statistical Education• Statistical Methodology• Survey Research Methods• Mathematical Statistics
In addition, the forum will organize discussion sessions onstatistics education and research with case studies to exchangeideas and experiences.
Sponsor/Co-Sponsors
• Renmin University of China• Peking University• National Bureau of Statistics of China• China Institute of Government Statistics• University of Connecticut• International Chinese Statistical Association• Capital University of Economics and Business• Central University of Finance and Economics
Logistics
Free banquet will be served during the conference. Help willbe provided on local hotel booking.
Registration
• Before March 1: Free (with abstract submitted).• After March 1: 800 RMB for domestic participants; 150
USD for overseas participants; 50% off for students.
Abstract Submission
Please submit abstract before April 30, 2014, at http://
stat.ruc.edu.cn/news/tongzhi/2013/1231/674.html.Potential North American participants please send in-
quiries to Jun Yan ([email protected]).
Membership Application & Renewal Form
Name (Last) (Middle) (First)
(English)
(Chinese)
Affiliation:
Gender (optional):
Address
Office Address:
City:
State: Zip Code: Country:
Email: Telephone: FAX:
Home Address:
City:
State: Zip Code: Country:
Email: Telephone: FAX:
Education
Degree: Year Graduated:
University:
Professional Occupation & Title
Occupation: Title:
Membership Fees Regular (US$40)
Student (US$10)
Permanent (US$600)
Spouse (50%)
Donations
Total Amount Paid: US$
Statistical Area of Interest (circle all applicable): A: Agriculture B: Business / Economics
C: Computing / Graphics D: Education E: Engineering F: Health Sciences
G: Probability H: Social Sciences I: Biostatistics N: Theory & Methodology
Would you like to join the Biometrics Section (please circle your answer)? Yes No
Please Make Check Payable to: I.C.S.A. Mail This Form & Fees to: ICSA c/o Shuyen Ho, Statistics and Programming GSK, GlaxoSmithKline, 5 Moore Drive, PO Box 13398, Research Triangle Park, NC 27709, USA
From the ICSA 2013
Conference in Hong
Kong
2014 Joint Applied Stascs Symposium of ICSA and KISS
June 15-18, Marriot Downtown Waterfront, Portland, Oregon, USA
The ICSA 2014 (23rd
) Applied Sta(s(cs Symposium will be held from June 15 (Sunday) to June 18 (Wednesday) at
the beau(ful Rose city of Portland, Oregon. This will be a joint conference with the Korean Interna(onal Sta(s(-
cal Society (KISS). The symposium will start with one day of short courses followed by scien(fic programs on 16 -
18th (Monday to Wednesday), and also includes student paper contests and social events.
Keynote speakers
♦ Dr. Sharon-Lise Normand, Professor, Departments of Health Care Policy, and Biosta(s(cs, Harvard University
♦ Dr. Robert Gentleman, Senior Director, Bioinforma(cs, Genentech
Honorable Banquet speaker
♦ Dr. Sastry Pantula, Dean, College of Science, Oregon State University & former president of American Sta(s(-
cal Associa(on (ASA)
Important Dates
♦ February 1, 2014: Symposium Registration, Hotel Reservation and Abstract Submission Open
♦ March 15, 2014: Deadline for Student Paper Award Applications
♦ April 1, 2014: Deadline for Online Registration and Abstract Submission (Invited Sessions)
♦ May 1, 2014: Deadline for Abstract Submission (Contributed Sessions)
♦ May 1, 2014: Deadline for Early Registration for Short Course and Symposium
Information about short courses, call for student paper contest and our committee members are described sepa-
rately in this issue of the ICSA Bulletin. More information could be found on our website http://www.statkiss.org/
icsakiss2014/. We have more than 60 invited sessions and will have more contributed sessions.
Portland is located near the confluence of the Willame?e and Columbia rivers with unique city culture. It is close
to the famous Columbia gorge, Oregon high mountains and coast. Oregon is also famous for many micro- brewer-
ies and beau(ful wineries, without sale tax. June will be a great (me to visit.
We are working hard to make this conference a memorable and worthwhile learning experience for all par(ci-
pants. We sincerely welcome all ICSA and KISS members, and all people interested in applied sta(s(cs to par(ci-
pate, submit abstracts to contributed sessions or provide sugges(ons. Please send any inquiries/sugges(ons to
Dongseok Choi or Rochelle Fu, [email protected].
The Executive Committee for 2014 ICSA-KISS Joint Applied Sta(s(cs Symposium
The city The city The city The city The city The city The city The city The city The city The city The city of roses of roses of roses of roses of roses of roses of roses of roses of roses of roses of roses of roses welcomes welcomes welcomes welcomes welcomes welcomes welcomes welcomes welcomes welcomes welcomes welcomes you in you in you in you in you in you in you in you in you in you in you in you in 2014!2014!2014!2014!2014!2014!2014!2014!2014!2014!2014!2014!
![oz 00 O o a O LU LU o z u.] d O a a z a U- CD o o o LU LU ... · z u.] d O a a z a U- CD o o o LU LU 1.14 LU o z LU o a > a U] a z o a o IL LU c-rj o o LU o a U) CD a LU a z o N o](https://img.pdfslide.us/doc/110x75/5f8b3edecf0c2b455c103fcd/oz-00-o-o-a-o-lu-lu-o-z-u-d-o-a-a-z-a-u-cd-o-o-o-lu-lu-z-u-d-o-a-a-z-a.jpg)
