myweb.facstaff.wwu.edumyweb.facstaff.wwu.edu/~chalmers/ElichMaterials/Elich_Course_… · Web viewKIN 506. Research Design in Human Movement and Performance. Lecture notes, Fall

KIN 506Research Design in

Human Movement and Performance

Lecture notes, Fall 2015Dr. Gordon Chalmers

Dept. of PEHRWestern Washington University

Ó 2015, Gordon Chalmers, Ph.D.

Thesis Format

The information below was copied from the Guidelines and regulations for masters degree candidates, downloaded from KPE program web site.

PROPOSAL FORMAT 1. Cover page containing the title of the proposal. 2. Chapter I - The Problem and Its Scope A. Introduction

B. Purpose of the studyC. Hypotheses or hypothesisD. Significance of the study E. Limitations of the studyF. Definition of terms

3. Chapter II - Review of Literature A. Brief introduction B. Review of pertinent literature - group according to commonality of topic or subtopic - synthesize and summarize each topic or subtopic C. Summary 4. Chapter III - Methods and Procedures A. Brief introduction B. Description of study population C. Design of study D. Data collection procedures 1. Instrumentation 2. Discussion of measurement techniques & procedures 3. Data processing or training program description (optional) E. Data (Statistical) AnalysisNote: The committee chair may request that a literature review table be constructed to summarize sections.

Each 506 student is to get a copy of a recent KIN program master degree thesis from the library. Choose a thesis that covers a topic area you are interested in potentially studying. E.g., exercise physiology, biomechanics, sport psychology etc. Below is a listing of some theses over the past several yearsByrkett, Jodie M. A comparison of the physical readiness testing scores of the United States Navy /

by Jodie M. Byrkett.2008. LD5778.9 .B965.Jackson, Morgan (Morgan Ada) Effect of inspiratory muscle training on respiratory muscle function,

exercise capacity, and quality of life in community dwelling elders / by Morgan Jackson.2008. LD5778.9 .J3326.

Brock, Benjamin. Effects of high and low velocity eccentric muscle actions on caloric expenditure and delayed onset muscle soreness / by Benjamin Brock.2008. LD5778.9 .B8626.

Schwerdtfeger, Katrina. Effects of inspiratory muscle training on arterial oxygen-hemoglobin saturation in female collegiate endurance runners / by Katrina Schwerdtfeger.2008. LD5778.9 .S35983.

Bies, Erik Stephen. Maximal exercise testing : the physiological responses of collegiate distance runners to treadmill and elliptical exercise machines. 2008. LD5778.9 .B563.

Sweeney, Kathryn. Motherhood and coaching burnout 2009. CALL # LD5778.9 .S95694.Wierzba, Jill. Psychological and emotional aspects of injury : rehabilitation among female

intercollegiate basketball players 2009. CALL # LD5778.9 .W49938.Barlond, P. (Paul Aaron) Effects of high resistance training on heart rate variability in older adults

2009. CALL # LD5778.9 .B32556.

KIN 506 Lecture Notes Ó 2015, Gordon Chalmers, Ph.D. 3 Updated 2/8/16

Thesis Format

Dudley, Jason. The interrater and intrarater reliability of the functional movement screen 2010. CALL # LD5778.9 .D7895.

Grambo, Laura B. Heavy elastic vs. white tape : the effect of ankle taping on ankle range of motion 2010. CALL # LD5778.9 .G7317.

Sweeny, Matt. Comparison of linear and daily undulating periodization in resistance training using simple measures of overreaching / by Matt Sweeny. 2010. CALL # LD5778.9 .S9569584.

Teichler, Liza S. The relationship between bat velocity, upper and lower extremity power and the rotational kinetic chain in NCAA Division II softball players 2010. CALL # LD5778.9 .T3872.

Broderick, Kelly M. (Kelly McGrorey) Length of exercise history and depressive symptoms in community dwelling older adults / by Kelly McGrorey Broderick. 2010. CALL # LD5778.9 .B86293.

Cronin, Kevin J. (Kevin James) The effects of training status and exercise intensity on plyometric exercise volume. 2010. CALL # LD5778.9 .C7726.

Hahn, Teresa J. Comparing the effects of inspiratory muscle training and core training on core muscle function 2010. CALL # LD5778.9 .H18986.

Newton, Carl Q. (Carl Quinn) Effects of fatigue on muscle activation and shock attenuation during barefoot running 2010. CALL # LD5778.9 .N4887.

Choudhari, Anuja. Effects of inspiratory muscle training on heart rate variability 2010. CALL # LD5778.9 . C479.

Spickler, Eric R. Effect of resistance training on body composition of persons with type II diabetes 2010. CALL # LD5778.9 . S7336.

Digital copies of the following works are available at the link provided.Cleveland, Megan A. (2011) Effect of core strength on long distance running performance

http://content.wwu.edu/cdm/ref/collection/theses/id/406Buckman, Rahmin, (2011) Effects of a goal setting program on the exercise commitment and

fitness levels of university students, http://content.wwu.edu/cdm/ref/collection/theses/id/427Rasnack, Catherine N. (2011) Effects of a life skills program on the social and academic

performance of freshman student-athletes http://content.wwu.edu/cdm/ref/collection/theses/id/447

McFadden, Courtenay, (2011) Effects of inspiratory muscle training on anaerobic power in trained cyclists, http://content.wwu.edu/cdm/ref/collection/theses/id/451

Jones, Kelly A., (2011) Influence of mental toughness on the performance of elite intercollegiate athletes, http://content.wwu.edu/cdm/ref/collection/theses/id/420

Zuleger, Brian M. (2011) Leadership characteristics of successful NCAA Division I track and field head coaches, http://content.wwu.edu/cdm/ref/collection/theses/id/432

Harrison, Alexander, (2011) Postactivation potentiation : predictors in NCAA Division II varsity track and field power athletes, http://content.wwu.edu/cdm/ref/collection/theses/id/433

Keller, Marc A., (2012) Biomechanical comparison between a baseball pitch and a first serve in tennis, http://content.wwu.edu/cdm/ref/collection/theses/id/564

Huntington, Summer, (2012) Exercise adherence in sedentary university employees after an 8-week web-based intervention, http://content.wwu.edu/cdm/ref/collection/theses/id/559

Aarseth, Lindsay M. (2013) Initial effects of Kinesio Tape on shoulder joint position sense at increasing elevations, http://content.wwu.edu/cdm/ref/collection/theses/id/596


http://content.wwu.edu/cdm/ref/collection/theses/id/596










http://content.wwu.edu/cdm/singleitem/collection/theses/id/406/rec/1

Academic Honesty Policies

Student self-study section on plagiarism and misinterpretation of the research articles

Appendix D of the WWU Catalog1 details the obligation of students and faculty to adhere to principles of academic integrity. Students must review this appendix and will be held to its requirements. These requirements apply not only to the thesis work, but to all academic work at WWU.

In order to ensure complete understanding of the requirements specifically regarding plagiarism, students are directed to the additional sources maintained by the WWU library2.

Some further clarification of the forms of plagiarism and breaches of academic integrity that are specific to our program are included below:

All statements of fact must be supported by references to the original source(s) of the information.

Students must conduct a literature search and retrieval in order to integrate the appropriate literature into their work. It is not acceptable to report previous results based only on the abstract. You must read and interpret the original journal article in order for it to be included in your work.

Students must accurately present the results and interpretations of their sources, which requires careful study of each source.

Students must provide complete and appropriate citations to the research literature, which includes the following:

o The student must provide a citation to the appropriate research for all information, thoughts, interpretations, and recommendations made by the authors of the research articles that the student wishes to include in his/her work. The citation must be in APA format, which includes citing the names of the authors and year of publication with the relevant statement.

o The student must present the information from the research articles in his/her own words. The student may not use the exact wording of the authors without also using quotation marks, but the use of quotation is scientific writing is very rarely, if ever done.

The information presented originates from the stated article, and therefore is a primary citation, and not a secondary citation. A secondary citation is when the information cited from an article (e.g. article A) is actually information presented in a previous article (B) and cited by the authors of article A. You must access article B in order to discuss their results and conclusions.

Faculty members will routinely conduct reference checks on student work to ensure that students are in compliance with these requirements.

1http://catalog.wwu.edu/content.php? catoid=5&navoid=312&returnto=search#Academic_Honesty_Policy_and_Procedures 2 http://libguides.wwu.edu/plagiarism and http://lib199.lib.wwu.edu/ref/subjects/govinfo/twain.htm


http://lib199.lib.wwu.edu/ref/subjects/govinfo/twain.htm

http://libguides.wwu.edu/plagiarism

http://catalog.wwu.edu/content.php?catoid=5&navoid=312&returnto=search#Academic_Honesty_Policy_and_Procedures

http://catalog.wwu.edu/content.php?catoid=5&navoid=312&returnto=search#Academic_Honesty_Policy_and_Procedures


Failure to comply with the requirements stated in 1) Appendix D, 2) the additional

sources provided in the links in the footnotes, or 3) this document would be a form of plagiarism or academic dishonesty, and appropriate disciplinary action will be taken, according to the policy set forth by the Catalog and including any additional requirements identified by the Graduate School and our program and department. Disciplinary action may include a letter to university officials detailing the infractions, delay of graduation, unwillingness of faculty members to serve on your committee, and may even result in dismissal from the program.

An exercise that demonstrates the appropriate use of the research literature

The goal of this exercise is to help you learn how to interpret and cite information in a research article. This assignment will ask you to recognize errors in the interpretation and citation of research articles. Your detailed attention to this now will pay off later when you complete your thesis and other academic work by helping you to recognize how to avoid problems with accidental plagiarism or misinterpretation of the research articles.

Instructions: Obtain the following article (consider saving paper by bringing it up on your screen and not printing it), and then study the questions and answers that follow to understand the differences between appropriate use of the research literature, and either academic dishonesty and/or plagiarism.

Wright R. W., Steger-May K., & Klein S. E. (2007). Radiographic findings in the shoulder and elbow of Major League Baseball pitchers. American Journal of Sports Medicine, 35, 1839-1843.

1. Which of the following represents an accurate depiction of this study’s results? A. Based on radiograph images, the pitching shoulder and elbow of major league pitchers

were found to exhibit signs of tissue degeneration, however, the detection of these degenerative changes in the preseason did not lead to the ability to predict which pitchers would become injured enough during the subsequent season to end up on the disabled list (Wright, Steger-May, & Klein, 2007).

B. The use of radiographic images during the preseason to detect shoulder tissue degeneration among major league pitchers is highly predictive of the time the pitcher will spend on the disabled list (Wright, Steger-May, & Klein, 2007).

C. Wright, Steger-May, and Klein (2007) determined that the degeneration of the tissues in the shoulder and elbow of major league pitchers had occurred because the pitchers had sustained acute injuries, and not, as had been hypothesized by other researchers, because of chronic repetitive stress.

D. The number of degenerative changes in the elbow that were apparent on a radiographic image were not related to the number of seasons or innings pitched (Wright, Steger-May, & Klein, 2007).

2. Which of the following represents an accurate depiction of this study’s results? A. Major league pitchers were found to have an average of 8.5 injuries in the shoulder, as

detected by preseason radiographs (Wright, Steger-May, Klein, 2007).



B. The most common injury detected in the elbow was a glenoid osteophyte (Wright, Steger-May, & Klein, 2007).

C. The most common injury detected in the shoulder was a glenoid osteophyte (Wright, Steger-May, & Klein, 2007).

D. The pitchers in this study were found to pitch an average of 8.5 innings per game (Wright, Steger-May, Klein, 2007).

3. Which of the following represents an accurate depiction of this study’s results? A. Acromioclavicular joint damage was not associated with the number of innings pitched

for major league baseball pitchers (Wright, Steger-May, & Klein, 2007). B. Due to an elbow injury, seven pitchers were placed on the disabled list for two

consecutive seasons (Wright, Steger-May, & Klein, 2007). C. Of the 57 pitchers included in the study, 14 spent time on the disabled list at some

point during the subsequent season due to a shoulder injury (Wright, Steger-May, & Klein, 2007).

D. Pitchers who were placed on the disabled list during the season were found to have statistically significantly more degenerative abnormalities in the elbow during the preseason than those who were not on the disabled list (Wright, Steger-May, & Klein, 2007).

4. Which of the following represents a “secondary citation”? (hint: examine the Introduction section)A. In Major League Baseball, shoulder and elbow injuries cause the most absences from

play than any other injuries (Wright, Steger-May, Klein, 2007). B. The analysis of radiographic images of the pitching shoulder and elbow revealed that

multiple sites of degeneration are commonly found in pitchers, and that increases in pitching time are associated with increased degeneration (Wright, Steger-May, & Klein, 2007).

C. The presence of degenerative changes on preseason images was associated with increased pitching time, but were not predictive of injury later in the season (Wright, Steger-May, & Klein, 2007).

5. Which of the following represents plagiarism? (hint: examine the Results section, 4th paragraph)A. The pitchers who ended up on the disabled list did not differ from their counterparts

when comparing age and number of seasons and innings pitched (Wright, Steger-May, & Klein, 2007).

B. Comparing pitchers on the disabled list for a shoulder injury with those not on the disabled list, there was no significant difference in the age, number of seasons pitched, or number of innings pitched between the 2 groups (Wright, Steger-May, & Klein, 2007).

C. There was no significant difference between pitchers on the disabled list and those who were not on the disabled list, based on age and experience (measured as number of seasons and innings pitched) (Wright, Steger-May, & Klein, 2007).



6. Which of the following represents the best way to phrase this information while appropriately citing the authors? (hint: examine the 4th paragraph in Discussion section)A. As with the studies mentioned above, multiple findings were noted on radiographs.

These findings, however, do not represent an acute injury. Rather, they are changes that occur as a result of the chronic, repetitive stresses incurred by pitching at an elite level. These radiographic findings did not predict future time on the disabled list (Wright, Steger-May, & Klein, 2007).

B. Many degenerative changes were detected in the preseason images of the pitchers’ shoulders and elbows, however, the changes were chronic - and not acute - in nature, and that they occurred because of the recurring stresses experienced by the shoulder and elbow during competitive pitching (Wright, Steger-May, & Klein, 2007). Further, the authors found that the degenerative transformations noted on the preseason images were not unique to players who would be on the disabled list during the subsequent season.

C. Many degenerative changes were detected in the preseason images of the pitchers’ shoulders and elbows, however, the changes, as noted by Wright, Steger-May, and Klein (2007) “do not represent an acute injury.” Instead, the authors determined that “they are changes that occur as a result of the chronic, repetitive stresses incurred by pitching at an elite level. These radiographic findings did not predict future time on the disabled list.”

Answers to the preceding questions: 1. ‘A’ represents the accurate description of results, and the other answer options are not the

results or implications of the study. 2. ‘C’ represents the accurate description of results, and the other answer options are not the

results or implications of the study. 3. ‘C’ represents the accurate description of results, and the other answer options are not the

results or implications of the study. 4. ‘A’ represents a secondary citation, because this information is presented in introduction,

is not a point under study in this article, and the authors cite the 4th reference for this information (Conte, Requa, & Garrick, 2001). The other answer options represent the conclusions of the study, appropriately paraphrased from the first and last paragraphs of the discussion, and so are appropriate uses of this article, and are not secondary citations.

5. ‘B’ is plagiarism, as it is the verbatim use of the authors’ words, without the use of quotation marks. The other answer options report on the same information as in the plagiarized version, but are paraphrased so that they do not constitute plagiarism.

6. ‘B’ represents paraphrasing without plagiarizing and without the over-use of quotes, whereas the other answer options either display an unnecessary use of quotes, which is not effective writing, or the verbatim use of the authors’ words with no quotes, which is plagiarism.


Unit 1: Introduction to Research

BASIC RESEARCH USUALLY: APPLIED RESEARCH USUALLY: theoretical problems animal subjects laboratory setting results have limited direct

application

address immediate problems use human subjects limited control over research

setting results are direct value to

practitioners

THE SCIENTIFIC METHOD OF PROBLEM SOLVING

Step1: Developing a problem (defining & delimiting it)

Identification of dependent & independent variables

INDEPENDENT VARIABLES (experimental or treatment variable)what the researcher is manipulating

DEPENDENT VARIABLESthe effect of the independent variable

Step 2: Formulating a hypothesis

hypothesis = expected resultbased on theories, previous studies, intuition

hypothesis must be TESTABLEsupported or refuted

hypothesis can not be a value judgment

Step 3: Gathering the data

Wish to maximize internal validity & external validity

INTERNAL VALIDITYthe extent to which the results can be attributed to the TREATMENTS used in the study

Controlling all other variables that can influence the results, so that you can say any result you get is due to the treatment can be difficult hard for you to think of everything to control when planning a study have as many people as possible read & critique your research design

EXTERNAL VALIDITYthe generalizability of the results of the study


Unit 1: Introduction to Research

the more you control (to increase the internal validity) the LESS generalizable the conclusion become

Step 4: Analyzing the datastatistics (easy once you learn the mechanics)but which statistic to use?

Step 5: Interpreting results

Easy (if you did steps 1-4 well)Deadly (if you did steps 1-4 poorly)

THE SCIENTIFIC METHOD IS NOT THE ONLY PROBLEM SOLVING METHOD case studies systematic observations descriptive research (questionnaires, interviews) historical research philosophical research research synthesis (e.g., Dr. Caine’s recent work) developmental studies (longitudinal methods)


Unit 2: Developing a Problem and Using the Literature

Problems to solve come from: theoretical frameworks (leads to basic research) practical problems to solve (leads to applied research)

TALK, TALK, TALK...• faculty• fellow grad students• advanced grad students

Look for... research being done (it leads to new questions) series of studies that are not completed

When you think you have an idea,Do a thorough literature review to:

• stimulate your ideas, help define your hypothesis• look for unanswered questions• find support for a hypothesis you have• understand methods used in your field• ensure that the question you are asking has not already been answered

LITERATURE SEARCH STRATEGIES

Start with SECONDARY sources= text books, review papers, encyclopedia

• gives a summary & evaluation of previous work in the field• seldom used as references in thesis

Move to PRIMARY sources= journal articles, thesis, dissertations (if not published as journal article)• gives details of studies• principal references in thesis

WRITE A PROBLEM STATEMENT

FIND SECONDARY SOURCES:catalogue (text books, encyclopedias)use search techniques for primary sources, but restrict search results to reviewsdetermine descriptors for search

SEARCH PRIMARY SOURCES:• indexes (Medline, sport discus, ERIC, PsycINFO, etc.)• bibliographies (including at the end of papers you read)

“why didn’t my computer search reveal that paper..?”

HOW TO READ RESEARCH: don’t look for eternal truths in each paper. Look for ideas and indications, patterns when papers are combined. No one paper proves anything.

read the abstract first, to get the overview



don’t read only the abstract, read for: well defined purpose (introduction) characteristics of the subjects instruments & tests used (demonstrations of validity & reliability?) independent & dependent variables treatments applied (if experimental study) research design statistical methods actual data (results) to draw your own conclusions (do not just read and

believe the authors conclusions) results implications of findings (discussion) questions raised for future study (ones they raise seldom stated, or ones you

see) citations to other important works

Summarize the results of each individual paper • Electronic notes added to pdf file? paper stapled to article? file folders for articles by themes?

The Literature Review (Thesis Chapter II)INTRODUCTION

purpose of review (& how?)must be quick and to the point

BODYorganized around important topics

synthesize the topic don’t restate the results, tell what can only be seen by looking at the collection of

articles, patterns, trends, missing information

BORING, BORING, BORING...study 1 descriptionstudy 2 description...study 30 description

BETTER:Present a concept, then discuss & cite papers that relate to this• some in detail• less important ones that give same conclusion combined into one sentence

Ensure you critically examine the theory, methods, author’s interpretations of studies

Don’t just say what they did & found!



Ensure your review is COMPLETE.



You should know ALL the related literature

SUMMARYOverall summary of the different topics, again tell what can only be seen by looking at the collection of articles

THE REVIEW MUST DEMONSTRATE THAT:• your problem needs investigating• you have taken previous work into account to develop your hypothesis• you know how previous work relates to your planned work


Unit 3: Presenting the Research Problem

PRESENTING THE RESEARCH PROBLEM(Your Chapter I)

(A) Introductionreasonably briefdo not be too technicalcreate interest in the topic

Some review is neededThe reader has not been living this topic

A well written introduction leads to the purpose / problemThe reader should be able to guess what these will say

Since the time of the ancient Greeks and Romans, the athletic arena has traditionally been considered a male domain. In the year 200, a Greek physician proclaimed that menstrual blood was accumulated fluids which resulted from an idle life, and by 1488, the public believed, due to the opinion of the Catholic church, menses was a disease (Weiss, 1997). Women have developed a great deal throughout the centuries. While women were first allowed to complete in the Olympic Games in 1912, and events such as the women’s marathon were only added in 1984 (Lebrun, 1993). Women have accomplished great physical advances over the past few decades, however, little research exists on the physiological effects of aerobic capacity in relation to hormonal status in trained athletes (Loucks, 1990).

Relatively little work has been done on the influence of hormones such as estrogen and progesterone on various components of athletic performance, however, from the research that has been done the results are contradictory (Lebrun, McKenzie, Prior, & Taunton, 1995). There is some evidence that menstrual cycle phase and the associated variations in the female steroid hormones effect athletic performance (Quadango, Faquin, Nam Lim, Kuminka, & Moffatt, 1991). Monthly cyclical changes such as basal and core body temperature, metabolic rate, perceived exertion, substrate metabolism, hemoglobin concentrations, heart rate, minute ventilation, respiratory exchange ratio, anaerobic performance, endurance time to fatigue, aerobic capacity, muscular strength are some of the areas that may be effected by hormonal fluctuations (Lebrun et al 1995).

There is a large amount of myth, anecdotal, and unscientific information about performance and female steroid hormones. There is controversy within the research community about the effects of hormones and performance in women (Nicklas, Hackneu, & Sharp, 1989; Stephenson, Kolka, & Wilkerson, 1982; Lamont, Lemon, & Bruot, 1987; Souza, Maguire, Rubin, & Maresh, 1990). A number of the studies looking at performance in relation to menstrual phase have used untrained athletes as subjects, while it has been suggested that the effects of hormones on performance may be too small to be detected in non-athletes (Lebrun et al 1995).



The 20th century has brought ever-increasing numbers of athletic women into the competitive area. This increase in participation by women has escalated the awareness of issues specific to active women. The suggestion that menstrual phase may effect the performance of trained athletes is provocative. Highly active women might be at a physiological advantage during certain phases of the month compared to other women at different phases.Stephanie Barss, PE 506 Fall 1997

(B) The Purposestated as “purpose” statement

IDENTIFY THE VARIABLES:• dependent• independent (treatment)

Categorical variable (if any) = independent variable that can not be manipulatede.g., age, race, gender, etc. differencesPurpose of the Study

The purpose of this study was to determine if a significant difference existed in maximal oxygen consumption in women in two difference phases of menstrual cycle. The two phases were the midfollicular and midluteal phases of the menstrual cycle.

(C) Hypothesisstated as Null hypothesis or the research hypothesis (one or the other)

Research hypothesis (H1)expected resultsyou expect that there will be a difference found between your experimental

groups

Null hypothesis (Ho)opposite of research hypothesisthat there is NO treatment effect, no difference between treatment groups

(D) Significance of the studyBasic research study:study significance is based on its ability to contribute to formulation or

validation of a theory

Applied research study:study significance is based on its ability to contribute to the solution of

some immediate problem

(E) Limitations of StudyDELIMITATION



= A limitation, imposed by the investigator, in the scope of the studya choice the researcher makes to effect a workable research problem•reduces the external validity

LIMITATION= A possible shortcoming or influence that either can not be controlled or

is the result of delimitations imposed by investigator• reduces the internal validity

(F) Definition of Termsoperational definition for your study

e.g., fatigue means?inability to continue at 80% max aerobic capacityinability to lift 20 lb weigh to full flexioninability to continue at 5 mph on treadmill


Unit 4: Planning and Reporting Your Methods

PLANNING & REPORTING YOUR METHODS(Chapter III)

A thesis has much more detail in methods than a published research article does

Much of the added detail in methods may be placed in the appendix of the thesis

General comment on methods/ research design:LESS IS MORE

(except in # of subjects)

A complex study with many different treatment groups (independent variables), and many dependent variables becomes difficult to run, control, recruit enough subjects for, statistically analyze.

METHODS SECTION

(A) Introduction

(B) Description of study population

Some parameters you may want to report: (but all not needed for all studies)

age gender trained - how is this defined? novice / untrained size (weight, height, body composition) special characteristics (athletes, students, runners, living in retirement

home) number of subjects

(C) Design of the study

A well designed study is one in which the only explanation for a change in the dependent variable(s) is how the subjects were treated (independent variable)

Discussed in unit 11

(D) Data collection procedures

Intro.: when, where, how much?

1. Instruments



discussion of equipment, standardized tests, etc.

2. Discussion of techniques & procedures specific order in which steps were taken exact method for how instruments were used timing of procedures instruction given safeguards how you will ensure subject compliance any other important details3. Data processing or training program description (optional)

Your goal is to ensure that another researcher reading the methods section can replicate the exact study

(E) Data (Statistical) Analysis

• type of statistical analysis• level of significance usedEtc.

Discussed in units 5-9

PILOT WORK USING YOUR METHODS

Whenever possible, do a pilot study on a small sample to ensure that your methods will work!



LAYOUT OF THESIS PROPOSALNote in this sample layout:

Single spaced to save space in demonstration, thesis is double spaced. There is no extra space placed before any of the section headings. This sample is for typical thesis requiring three levels of headings, see pg 62 APA Guide 6th

edition. For three levels, you use heading levels 1, 2, 3 See sample paper in 6th ed APA guide, page 41 – 51 To use four levels of headings if needed, see pg 62 of APA guide, 6th edition. References start on a new page PEHR thesis guidelines include that you must omit the running header specified in the APA

guidelines. APA style for heading has changed from version 5 to 6 of APA. The style below is 6th

edition.

Chapter IThe Problem and Its Scope

IntroductionBlaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff

more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff.Purpose of the Study

Blaa blaa blaa WHAT you will do stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff.Hypotheses

Blaa blaa blaa stuff Hypothesis (if singular) stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff.Significance of the Study

Blaa blaa WHY you will do the study blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff.Limitations of the Study 1. List2. List3. ListDefinition of TermsListListList

Chapter IIReview of Literature

IntroductionBlaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff

more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff.Review of the Pertinent Literature

First subtopic within review. Blaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff ..

Second subtopic within review. Blaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff ..


Heading level 1 – Centered, boldface, Uppercase and Lowercase, use roman numerals, laid out on two lines

Heading level 2 – Flush left, Boldface, Uppercase and Lowercase

Heading level 3 – Indented, boldface, lowercase paragraph heading (only first letter of the first word is uppercase, unless a proper noun is included) ending with a period. Text follows immediately on same line.


SummaryBlaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff

more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff.Chapter III

Methods and ProceduresIntroduction

Blaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff more stuff.Description of Study Population

Blaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff more stuff.Design of the Study

Blaa blaa blaa Use formal names of study designs discussed in unit 11 stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa.Data Collection Procedures

Instrumentation. Blaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff more stuff.

Measurement techniques and procedures. Blaa blaa blaa stuff stuff stuff more stuff more stuff more. Blaa blaa blaa stuff stuff stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff more stuff.

Data processing. Blaa blaa blaa not statistics, calculation procedures on values obtained from instruments stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff.

Training procedures. Blaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff more stuff.Data Analysis

Blaa blaa blaa statistical analysis- remember level of sig (+ Bonferonni correction?), software (including version) used stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff more stuff more stuff more stuff. Blaa blaa blaa stuff stuff stuff more stuff.

References


Heading level 3 – Only one capital at start

Heading level 1, references start on new page

Unit 5: Advanced correlation & regression topics

Interpretation of strength of r:± 0.80 - 1.00 Very Strong relationship± 0.60 - 0.79 Strong relationship± 0.40 - 0.59 Moderate relationship± 0.20 - 0.39 Weak relationship± 0.00 - 0.19 No relationship

there is no set rule for interpretation, these cutoffs from: Salkind, N. J. (2000) Statistics for people who (think they) hate statistics. Thousand Oaks, CA: SageOR

Correlation Coefficient Descriptor 0.0-0.1 trivial, very small, insubstantial, tiny, practically zero 0.1-0.3 small, low, minor 0.3-0.5 moderate, medium 0.5-0.7 large, high, major 0.7-0.9 very large, very high, huge 0.9-1 nearly, practically, or almost: perfect, distinct, infinite Source: A New View of Statistics, http://www.sportsci.org/resource/stats/index.html

YOU CAN DO A CALCULATION OF THE SIGNIFICANCE OF A CORRELATION SCORE (often omitted)

Table A.2 Appendix A VincentTable A.3 Appendix A T & N

df = degrees of freedom = Npairs – 2

for 95% confidence use middle 0.05 column:Note: round down to the next available entry in the tableif |absolute value of r| > column value

result is significant at the p value of the column

e.g., 10 pairs r = -0.92df = 8middle column value = 0.632| -0.92| >0.632correlation is significant p < 0.05


Unit 6: Comparisons of two groups: effect size, power, sample size, Confidence Intervals

WHAT DOES LEVEL OF SIGNIFICANCE REALLY MEAN?

= PROBABILITY OF REJECTING THE NULL HYPOTHESIS WHEN IT IS TRUE (type I error).alpha = 0.05 means your decision to reject the null hypothesis will be incorrect 5% of the timei.e., 1 in 20 t-tests you do will falsely reject the null.

BUT:Is a significant difference meaningful?

USE: EFFECT SIZE (ES)

ES = |(M1 - M2) / s|(Called “Cohen’s d” in psychology stats books, but “effect size” in exercise science literature)

M1 = mean group 1M2 = mean group 2s = standard deviation control group

or pooled standard deviation of both, Thomas & Nelson

pooled standard deviation = sp

sp = s1

2 n1 1 s22 n2 1

n1 n2 2

s12 = variance in group 1

s22 = variance in group 2

n1 = # subjects group 1n2 = # subjects group 2


NO DIFF DIFF

NO DIFF

DIFF

YOUR CONCLUSION


NOTE: This ES calculation works for both paired (repeated measures) and two-group T-tests3, 4, see Unit 6 appendix at end of unit 6, although some articles state you must use a different formula for ES for paired t-tests 5

E.G., WEIGHT LOSS STUDY IN MEN.Q: Can you use the results of this study same to say that the weight loss program will be effective for men of the population the sample was drawn from?

i.e., was there a significant change in weight, at the alpha = 0.05 level?

MAN Jan Jun1 189 1722 201 1953 225 2094 195 1965 192 1856 203 2007 210 2058 241 2309 196 18910 188 19011 179 185

MEAN 201.7 196.0s 17.9 15.3

Using EXCEL, t-Test: Paired Two-Sample for Means

Variable 1 Variable 2 Mean 201.727273 196Variance 319.418182 232.6Observations 11 11Pearson Correlation 0.91828222Pooled Variance 250.3Hypothesized Mean Difference 0df 10t 2.64902373P(T<=t) one-tail 0.01217692t Critical one-tail 1.81246151P(T<=t) two-tail 0.02435383t Critical two-tail 2.22813924

for dependent t-test weight loss example:

3 Hojat, M. Xu, G., A Visitor's Guide to Effect Sizes - Statistical Significance Versus Practical (Clinical) Importance of Research Findings, Advances In Health Sciences Education, 2004, Vol 9; Number 3, pages 241-2494 Dunlop, W. P., Cortina, J. M., Vaslow, J. B., & Burke, M. J. (1996). Meta-analysis of experiments with matched groups or repeated measures designs. Psychological Methods, 1, 170-177.5 Joe W. Kotrlik, Heather A. Williams (2003) The Incorporation of Effect Size in Information Technology, Learning, and Performance Research Information Technology, Learning, and Performance Journal, Vol. 21, No. 1, 1-7



ES = |(M1 - M2) / s|= |(201.7 - 196.0) / 17.9| = 0.32

Note: Using s of the pre-program for the control s = 319 = 17.9

Interpretation of ES

<0.2 none 0.2 – 0.49 Small 0.5 – 0.79 Moderate0.8 LargeCohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). New Jersey: Lawrence Erlbaum. (cited in Hopkins, W. G. (1997), & Vincent (1999))

<0.2 Trivial 0.2 – 0.59 Small 0.6 – 1.19 Moderate1.2 – 1.9 Large 2.0 – 3.9 Very Large 4.0 Nearly perfectHopkins, W. G. (1997) A scale of magnitudes for effect statistics [Web Page]. URL http://www.sportsci.org/resource/stats/index.html [Accessed: 2001, November 29].

See also: Will Hopkins, How to Interpret Changes in an Athletic Performance Test., SportScience.org, Vol 8, 2004, Available: http://sportsci.org/jour/04/wghtests.htm, Retrieved 12-8-04.

Rhea MR. Determining the magnitude of treatment effects in strength trainingresearch through the use of the effect size. J Strength Cond Res. 2004Nov;18(4):918-20.

“Proposed labels are a conventional frame of reference…An effect considered small of trivial in one area may be substantial in another area” (pg 54, Ottenbacher, K. J., & Barrett, K. A. (1989). Measures of effect size in the reporting of rehabilitation research. Am J Phys Med Rehabil, 68(2), 52-8.

Another method to measure the size of the effect:2 = % of the difference between the groups attributable to the independent variable you are testing (pg 131 Vincent)



e.g., % of difference in strength due to age % of difference in aerobic capacity due to nutrient supplement % of difference in performance due to psychological treatment

2 = |(t2 -1) / (t2 + N1 + N2 -1)|N1 = # subjects in group 1 N2 = # subjects in group 2

for dependent t-test weight loss example:2 = |(t2 -1) / (t2 + N1 + N2 -1)|t = 2.65 N1 = 11 N2 = 112 = |(2.652 -1) / (2.652 + 11 +11 -1)| = 0.2525% of the difference in weight is attributed to the treatment



Is inclusion of Effect Size necessary in your data analysis?

Australian Conference of Science and Medicine in Sport 1998 Adelaide 13-16 October 1998

Free Paper & Poster Abstracts

Power, effect size and the misinterpretation of statistical inference in exercise science

H Speed*1 & M Andersen2

1School of Human Movement, Deakin University, Melbourne, Australia2Centre for Rehabilitation, Exercise and Sport Science, Victoria University, Australia

Past research into the power of research designs in medicine, psychology, education and the sports sciences (Christensen & Christensen, 1977) has revealed that low statistical power is a ubiquitous phenomenon across many fields of inquiry. The power of research designs in studies published in the 1995 volume of the Australian Journal of Science and Medicine in Sport (AJSMS) were analysed for their ability to detect small, medium and large effects according to Cohen’s conventions. Also examined were the reporting and interpretation of effect sizes along with experiment-wise error (alpha slippage, increased probability of Type I error) and the misinterpretation of nonsignificant results (possible Type II errors). The median power of the studies to detect small, medium and large effects were .10, .46 and .84, respectively. These results suggest that exercise science, at least as represented in the AJSMS is woefully underpowered and essentially incapable of detecting small, but possibly meaningful effects, has less than a coin toss of a chance detecting medium size effects, and has only adequate power to detect large effects. Not once, except in post hoc analyses, did the researchers control for experiment-wise Type I error rate. One study actually had a probability of Type I error of .97 due to lack of adjustment for performing 72 statistical tests. The reporting of effect sizes was nonexistent. Also, many studies reported only P values and not the magnitude of the test statistic, making it impossible to estimate effect sizes. In many studies with low power, nonsignificant results were interpreted as support for the null hypothesis. Thus, the likelihood of Type II errors in studies published in AJSMS is substantial. What exercise science researchers do not appear to understand is that effect size is the statistic of interest in research, not the highly manipulable P value, and that claiming "no difference" between groups in a study with questionable power should not be made unless effect sizes have been considered and a power analysis has been conducted. An appeal is made for exercise scientists to examine, report and interpret effect sizes rather than solely rely on P values to determine whether significant changes occurred or significant relationships exist.

References:Christensen, J. E., & Christensen, C. E. (1977). Statistical power analysis of health, physical education, and recreation research. Research Quarterly, 48: 204-208.

Speed, H, Power, effect size and the misinterpretation of statistical inference in exercise science – abstract. In, Australian Conference of Science and Medicine in Sport, Adelaide Convention Centre, Adelaide 13-16 October 1998 : abstracts, Canberra, Sports Medicine Australia, 1998, p.215. Available: URL: http://www.ausport.gov.au/fulltext/1998/acsm/smabs215.htm

Zhang, L; Qi, G, Effect size, a data analysis index to which much importance should be attached in exercise psychological research. Journal of Beijing University of Physical Education (Beijing) 21(1), 1998, 13-18.



Thomas, J R; Salazar, W; Landers, D M, What is missing in p is less than .05? Effect size. Research quarterly for exercise and sport (Reston, Va.) 62(3), Sept 1991, 344-348.

Uniform Requirements for Manuscripts Submitted to Biomedical Journals: Writing and Editing for Biomedical Publication

Updated February 2006http://www.icmje.org/#prepareIV.A.6.c. Statistics

Describe statistical methods with enough detail to enable a knowledgeable reader with access to the original data to verify the reported results. When possible, quantify findings and present them with appropriate indicators of measurement error or uncertainty (such as confidence intervals). Avoid relying solely on statistical hypothesis testing, such as the use of P values, which fails to convey important information about effect size. (Bold added by Chalmers) References for the design of the study and statistical methods should be to standard works when possible (with pages stated). Define statistical terms, abbreviations, and most symbols. Specify the computer software used.

Discussion: Each student tells 2 things they learned from the article:

Sifting the evidence what's wrong with significance tests? •Jonathan A C Sterne, George Davey Smith, BMJ 2001; 322: 226-231. Available: http://bmj.com/cgi/reprint/322/7280/226

Also available from Index line of course web page


http://bmj.com/cgi/reprint/322/7280/226

http://www.icmje.org/index.html


Ioannidis JPA. Why Most Published Research Findings Are False (2005). PLoS Medicine Vol. 2, No. 8, e124 doi:10.1371/journal.pmed.0020124

Available: http://medicine.plosjournals.org/perlserv?request=get-document&doi=10.1371%2Fjournal.pmed.0020124

Section of introduction:

Several methodologists have pointed out that the high rate of nonreplication (lack of confirmation) of research discoveries is a consequence of the convenient, yet ill-founded strategy of claiming conclusive research findings solely on the basis of a single study assessed by formal statistical significance, typically for a p-value less than 0.05. Research is not most appropriately represented and summarized by p-values, but, unfortunately, there is a widespread notion that medical research articles should be interpreted based only on p-values. Research findings are defined here as any relationship reaching formal statistical significance, e.g., effective interventions, informative predictors, risk factors, or associations. “Negative” research is also very useful. “Negative” is actually a misnomer, and the misinterpretation is widespread. However, here we will target relationships that investigators claim exist, rather than null findings.


http://medicine.plosjournals.org/perlserv?request=get-document&doi=10.1371%2Fjournal.pmed.0020124

http://medicine.plosjournals.org/perlserv?request=get-document&doi=10.1371%2Fjournal.pmed.0020124


Example of data collection and interpretation.

Shoulder ShoulderControl TreatmentGroup GroupStarting Angle

Starting Angle

136.3 92.0 E 55124.8 121.9 E 54114.6 143.5 E 53143.2 145.7 E 52129.1 95.3 E 51132.8 119.2 E 50132.3 131.7 E 49146.2 E 48137.6 148.5 E 47129.1 148.0 E 46 118.3 E 45136.3 92.0 D 44124.8 121.9 D 43114.6 143.5 D 42143.2 145.7 D 41129.1 95.3 D 40132.8 119.2 D 39132.3 131.7 D 38146.2 D 37137.6 148.5 D 36129.1 148.0 D 35 118.3 D 34136.3 92.0 C 33124.8 121.9 C 32114.6 143.5 C 31143.2 145.7 C 30129.1 95.3 C 29132.8 119.2 C 28132.3 131.7 C 27146.2 C 26137.6 148.5 C 25

148.0 C 24 118.3 C 23

92.0 B 22124.8 B 21114.6 143.5 B 20143.2 145.7 B 19129.1 95.3 B 18132.8 119.2 B 17132.3 131.7 B 16146.2 117.0 B 15137.6 148.5 B 14

148.0 B 13



118.3 B 12136.3 92.0 A 11124.8 121.9 A 10114.6 143.5 A 9143.2 145.7 A 8129.1 95.3 A 7132.8 119.2 A 6132.3 131.7 A 5146.2 117.0 A 4137.6 148.5 A 3129.1 148.0 A 2

118.3 A 1

MEAN 132.6 125.6 S 9.1 20.0 A data Treatment mean as % of control mean 94.7% ES 0.77 Moderate-difference between control and treatment mean angles Two group T-test P-value 0.321 MEAN 132.6 125.7 S 9.3 20.0

A & B data

Treatment mean as % of control mean 94.8% ES 0.74 Moderate-difference between control and treatment mean angles Two group T-test P-value 0.190 MEAN 132.7 125.9 S 9.2 20.0 A, B & C data Treatment mean as % of control mean 94.9% ES 0.74 Moderate-difference between control and treatment mean angles Two group T-test P-value 0.111 MEAN 132.7 126.1 S 9.0 19.9 A, B, C & D data Treatment mean as % of control mean 95.0% ES 0.73 Moderate-difference between control and treatment mean angles Two group T-test P-value 0.067 MEAN 132.7 126.1 S 9.0 19.9 A, B, C, D & E data Treatment mean as % of control mean 95.1% ES 0.73 Moderate-difference between control and treatment mean angles Two group T-test P-value 0.041



David Trafimow & Michael Marks (2015) Editorial, Basic and Applied Social Psychology, 37:1, 1-2, DOI: 10.1080/01973533.2015.1012991



POWER = Ability To Correctly Reject A False Null Hypothesis (i.e., to detect true differences between groups

= probability of making a correct decision)Desired power is 0.8 (80%) - POWER OF YOUR TESTS MUST BE REPORTED (Thomas & Nelson,

Vincent)

If power is too high: you wasted time and money doing excessive data collection.If power is too low: if you fail to reject the null, the failure to reject may be because you didn't have

enough power.

Power depends on: sample size & alphaWe usually don’t have control over alpha, the bigger the sample, the more powerful.

SAMPLE SIZE CALCULATION:You want to collect enough data (subjects) to have sufficient power, no more.

Determination of sample size depends on:a) alpha (0.05)b) desired power (0.8)c) difference in means between the two groups you are comparingd) the standard deviation within each of the two groups you are comparing

BUT: knowledge of c & d requires pilot data, or values obtained from previous similar research

If you know the required information sample size (to achieve desired power) may be calculated using software:

G*Power - available free at: http://www.psycho.uni-duesseldorf.de/abteilungen/aap/gpower3/or google search “gpower” to find current location for download

• sample must be randomly selected from populations which are normally distributed and have equal s

• all members must have an equal opportunity to be selected

true random sampling is seldom possible in our studies


http://www.psycho.uni-duesseldorf.de/abteilungen/aap/gpower3/


IDEALLY: Before you run experiment:

Estimate the needed sample size – using desired power, alpha, estimates of mean and variability

Example:What sample size is needed to detect a decrease in supine systolic blood pressure in elderly subjects?

Q1 – How big, or small, of a difference are you interested in detecting?e.g., is a 1% change important and you wish to be able to

detect this?

ANS: Lets assume a 10% change is important to you, anything less is not worth the time of the exercise program.

Solution:Use data from Chylla thesis, table4, control elderly supine

systolic BP = 142 18

Therefore: you treated group must have a value 10% less than thisi.e., 127 18 (assuming equal variability)

Put this data into nQuery advisor:

Answer: You need 24 subjects per group



Sample size calculation: further discussionUsing: http://calculators.stat.ucla.edu/powercalc/

1) Using same data as Chylla thesis data above in this unit. n= 48

Power = 0.807,

2) Now: assume the researcher can study 20 more subjects, does this increase the power of the study?

Scenario 1: add the additional 20 subjects evenly to both groups: n=68

Observe. Power = 0.923



Scenario 2: add all the additional 20 subjects to the experimental group#2, none to the control group: n=68

Observe. Power = 0.899To get a Power= 0.898, you only need 62 subjects, if you balance them across the groups.



Extreme example: Thesis project involving 12 treated subjects and 2 control subjects (not Chylla thesis data, but that data is used here to illustrate effect of very uneven group sizes) (small sample size of control group also violates assumptions of variances of groups since you can’t assume that the 2 control subjects provide a valid estimation of the control group variance)

. Does this study provide an acceptable level of statistical power?

Lesson to learn…. You get the most statistical power for a given # of subjects if you have even group sizes (if you have uneven group sizes you are wasting some of your time).

After you run experiment:Calculate the power your test had, using alpha, actual of mean and variability and sample sizes, and use this power value to help you interpret the results.

Interpretation of results of sig & power:

View 1 on post-hoc power analysis with your statistics:If you found no sig difference in your exp:A) if your power was high (i.e., >= 80%)

You can be fairly confident that there really is no sig difference between the groups.

B) if your power was low (i.e., < 80%)The lack of a sig difference between the groups may be because you didn’t have enough power to detect it (i.e., there is a sig difference, but you couldn’t detect it). You can not be fairly confident that there really is no sig difference between the groups.



e.g.:

View 2 on post-hoc power analysis with your statistics:Post-hoc power analysis does not provide you with any useful information, and should not be done

See: Hoenig, J., & Heisey, D. (2001). The abuse of power: The pervasive fallacy of

power calculations for data analysis. The American Statistician, 55, 19-24. Available: http://www.ac.wwu.edu/~chalmers/power.pdf

Discovering Statistics using SPSS for Windows, (2000) Andy Field, Sage, London., p 299

Hopkins, W., Stats queries: power, error inflation, Cronbach alpha. ACSM Biostats Interest group, on Sport Sci discussion group, Message 2549, http://groups.yahoo.com/group/sportscience/

“Don't do it as a so-called post-hoc power analysis. You can find heaps of references in papers and on the Web stating that post-hoc power analyses are misleading and should not be done. ”

Faul F, Erdfelder E, Lang AG, Buchner A. G*Power 3: a flexible statistical power analysis program for the social,behavioral, and biomedical sciences. Behav Res Methods. 2007 May;39(2):175-91.

THEREFORE: YOU MUST REPORT p (sig), an ES measure (ES or 2) (and

POWER? Depends on supervisor) for any stats test you run.

Do not think a “sig” or “non-sig” p-value determines the success or failure of an experiment

Interpret the combination of p, ES, (power), to interpret the results of your study.


http://www.ac.wwu.edu/~chalmers/power.pdf


Example of the use of both p-value and effect size to aid in the interpretation of experimental results.

Are the benefits worth the effort?

Table 1 ResultsTheir results in outlined box. Does this convince you to adopt the exercise program?

test 1 test 2

mean s mean s% change

sig (p<0.002) ES Cohen Hopkins

Get up and go 7.92 2.1 7.25 2 -8.5% y 0.319 small smallSit to stand 0.78 0.7 0.74 1.5 -5.1% 0.062 none trivialStanding reach (m) 0.3432 0.1 0.386 0.06 12.5% y 0.535 moderate smallTimed walk 7.5 1.8 8.1 2.2 8.0% 0.333 small smallSit and reach (PI) 0.248 0.1 0.289 0.13 16.5% y 0.293 small smallPhysical functioning 66.5 23 72.7 20.7 9.3% y 0.267 small smallRole physical 67.2 39 73.9 34.9 10.0% 0.174 none trivialPain 66.8 21 73.2 20.9 9.6% y 0.303 small smallGeneral health 68.9 19 70.1 17.2 1.7% 0.065 none trivialVitality 57.5 21 63.5 20 10.4% y 0.280 small smallSocial functioning 85.3 19 90 17 5.5% 0.242 small smallRole emotional 79.2 35 83.1 30.3 4.9% 0.111 none trivialMental health 76 15 79.5 16.2 4.6% 0.229 small smallFalls 0.18 0.6 0.08 0.3 -55.6% 0.156 none trivialEmergency department visits 0.11 0.3 0.03 0.18 -72.7% 0.235 small smallHospitalizations 0.08 0.3 0.01 0.09 -87.5% 0.259 small smallDoctor visits 1.34 1.1 0.91 0.87 -32.1% y 0.391 small smallNumber of medications 3.87 2.6 3.68 2.8 -4.9% 0.073 none trivial

What does the ES tell you about the effectiveness of the training program?



In excel 0.0002051 is written as 2.051E-4stats terminology excel terminology &

menu choiceSPSS terminology & menu choice

Dependent t-test T test: paired two sample for means

Paired samples t-test

Independent t-test T-test: two sample assuming equal variances

Independent samples t-test

NOTES:1. SPSS used "sig" instead of "P"2. As of fall 2009, SPSS software is called PASW (But may be referred to as SPSS)

Generally do a 2-tail test- because if there is going to be a difference found, you don’t know which group will be greater



Source: Clinical Gait Analysis Web Page http://www.univie.ac.at/cga/faq/statistics.html Retrieved 6-2-04

…Even when you have enough subjects for the various criteria to be satisfied (e.g. normal distributions) conventional (Fischer) statistics can still be quite misleading. If you speak to mathematicians these days they will often laugh when you mention Fischer and tell you that the only reason he did stats this way was because of the limitation in calculating power at the time. Modern statisticians are much more interested in computer simulation studies, which are apparently much more informative.

Dr. Chris Kirtley MD PhD Associate Professor, Dept. of Biomedical Engineering, Catholic University of America 620 Michigan Ave NE, Washington, DC 20064

….Another hang-up of Stern's, which is partially related, and reasonably well supported in the literature is to focus more on confidence limits in interpreting data rather than p-values.

…..

This whole area is a can of worms but you've got no option but to get to grip with it if you want to valid science.

Hope this is useful.

Richard Baker

Gait Analysis Service Manager, Royal Children's Hospital Flemington Road, Parkville, Victoria 3052 Adjunct Associate Professor, Physiotherapy, La Trobe University Honorary Senior Fellow, Mecahnical and Manufacturing Engineering, Melbourne University

High-intensity cycle interval training improves cycling and running performance in triathletes. Etxebarria N, Anson JM, Pyne DB, Ferguson RA., Eur J Sport Sci. 2013 Nov 9. [Epub ahead of print]Abstract Effective cycle training for triathlon is a challenge for coaches. We compared the effects of two variants of cycle high-intensity interval training (HIT) on triathlon-specific cycling and running. Fourteen moderately-trained male triathletes ([Formula: see text]O2peak 58.7 ± 8.1 mL kg-1 min-1; mean ± SD) completed on separate occasions a maximal incremental test ([Formula: see text]O2peak and maximal aerobic power), 16 × 20 s cycle sprints and a 1-h triathlon-specific cycle followed immediately by a 5 km run time trial. Participants were then pair-matched and assigned randomly to either a long high-intensity interval training (LONG) (6-8 × 5 min efforts) or short high-intensity interval training (SHORT) (9-11 × 10, 20 and 40 s efforts) HIT cycle training intervention. Six training sessions were completed over 3 weeks before participants repeated the baseline testing. Both groups had an ∼7% increase in [Formula: see text]O2peak (SHORT 7.3%, ±4.6%; mean, ±90% confidence limits; LONG 7.5%, ±1.7%). There was a moderate improvement in mean power for both the SHORT (10.3%, ±4.4%) and LONG (10.7%, ±6.8%) groups during the last eight 20-s sprints. There was a small to moderate decrease in heart rate, blood lactate and perceived exertion in both groups during the 1-h triathlon-specific cycling but only the LONG group had


http://www.ncbi.nlm.nih.gov/pubmed/24206175


http://www.ncbi.nlm.nih.gov/pubmed?term=Ferguson%20RA%5BAuthor%5D&cauthor=true&cauthor_uid=24206175

http://www.ncbi.nlm.nih.gov/pubmed?term=Pyne%20DB%5BAuthor%5D&cauthor=true&cauthor_uid=24206175

http://www.ncbi.nlm.nih.gov/pubmed?term=Anson%20JM%5BAuthor%5D&cauthor=true&cauthor_uid=24206175

http://www.ncbi.nlm.nih.gov/pubmed?term=Etxebarria%20N%5BAuthor%5D&cauthor=true&cauthor_uid=24206175

mailto:[email protected]

http://www.univie.ac.at/cga/index.htmlRetrieved%206-2-04


a substantial decrease in the subsequent 5-km run time (64, ±59 s). Moderately-trained triathletes should use both short and long high-intensity intervals to improve cycling physiology and performance. Longer 5-min intervals on the bike are more likely to benefit 5 km running performance.

Ankle Kinematics and Muscle Activity in Functional Ankle Instability. Monteleone BJ, Ronsky JL, Meeuwisse WH, Zernicke RF., Clin J Sport Med. 2013 Nov 13. [Epub ahead of print]Abstract OBJECTIVE: Following an ankle injury, many patients have functional ankle instability (FAI) with an increased predisposition to reinjury. The purpose of this study was to assess the effects of FAI on ankle kinematics and muscle activity during a lateral hop movement.DESIGN: Cross-sectional and observational study; all data collection for each subject was performed on 1 day.SETTING: Clinical biomechanics laboratory.PATIENTS: Two groups were studied: (1) Control group-no ankle injury (n = 12) and (2) FAI group (n = 12).INTERVENTIONS: The lateral hop movement consisted of multiple lateral and medial 1-legged hops over an obstacle (width, 72.5 cm; depth, 25.5 cm; height, 14.3 cm) onto adjacent force platforms. Each subject was instructed to perform as many lateral hops as possible during the 6-second trial. Means, SDs, 95% confidence intervals of the differences, and P-values were calculated.MAIN OUTCOME MEASURES: Ankle kinematics and muscle activity throughout the lateral hop movement.RESULTS: Significant differences existed between groups for mean (SD) dorsiflexion ankle positions-FAI 82.4 degrees (6.4) versus normal 75.2 degrees (10.1) and tibialis anterior normalized muscle activity-FAI 0.27 (0.21) versus normal 0.16 (0.13) at ground contact.CONCLUSIONS: The FAI group revealed greater tibialis anterior muscle activity and dorsiflexion ankle position at contact moving in the lateral direction. These differences between groups may have been related to an inherent predisposition to ankle injuries, a preexisting difference in task performance, a consequence of injuries, or a compensatory adaptation to previous injuries.

CONFIDENCE INTERVALS



http://www.ncbi.nlm.nih.gov/pubmed?term=Zernicke%20RF%5BAuthor%5D&cauthor=true&cauthor_uid=24231927

http://www.ncbi.nlm.nih.gov/pubmed?term=Meeuwisse%20WH%5BAuthor%5D&cauthor=true&cauthor_uid=24231927

http://www.ncbi.nlm.nih.gov/pubmed?term=Ronsky%20JL%5BAuthor%5D&cauthor=true&cauthor_uid=24231927

http://www.ncbi.nlm.nih.gov/pubmed?term=Monteleone%20BJ%5BAuthor%5D&cauthor=true&cauthor_uid=24231927



Population of all WWU students. Actual mean of this population = 150 lbs


The mean of each sample lies somewhere around, or occasionally, on the mean of the population

http://www.southalabama.edu/coe/bset/johnson/lectures/lec16_files/image006.jpg

COMPARE these sample statements in a results section:Less commonly seen in papers Most commonly seen in papers“…GRF was 28 ± 3 (mean ± 95% CI) N …” VERSUS “…GRF was 28 ± 2 (mean ± s) N …”

Tells:The range (25-31 N) that the POPULATION MEAN likely lies in (in 95% of cases in which as sample is taken)

Tells:Mean (28 N) and variability (s) of the SAMPLE

You REALLY want to learn about what the POPULATION is likee.g., all the people who are oldall the cyclists who use creatineall the people with the mental characteristics you examine

BUT: you don’t really care as much about your SAMPLE (except to ensure it does not appear strange), as you care about the POPULATION you are trying to learn about.

To calculate CI from sample:



Lower bound of 95% confidence interval = mean – (1.96 x SE)Upper bound of 95% confidence interval = mean + (1.96 x SE)(the derivation of this formula using z-scores is available in many statistic text books)

SE = standard deviation of the sample means (i.e. the variability of the different samples) is estimated by the following equationSE = s / √N

s = standard deviation of the sampleN= number of data points in the sample

If CI is small: true mean is close to the sample meanIf CI is large: true mean can be very different from the sample mean

The above is for a large sample, above about 30 (Field text, pg 42, 45)

For small sample CI:Lower bound of 95% confidence interval = mean – (tn-1 x SE)Upper bound of 95% confidence interval = mean + (tn-1 x SE)

Find t-value for a 2-tailed test with probability of 0.05

We wish to be able to say, with a certain degree of confidence (e.g., 95%) the range within which the true mean of the population lies (even though we can never know the true mean of the population because we can’t measure the weight of all the students, we can only take a sample of students).

i.e. 95% of the time we make a sample, the true mean of the population lies within the 95% confidence interval of the sample. (This is different from saying “I am 95% confident that the true mean lies within my 95% CI)

Details:One cannot say: "with probability (1 − α) the parameter μ lies in the confidence interval." One only knows that by repetition in 100(1 − α) % of the cases, μ will be in the calculated interval. In 100α% of the cases however it does not. And unfortunately one does not know in which of the cases this happens. That is why one can say: "with confidence level 100(1 − α) %, μ lies in the confidence interval." (source: http://en.wikipedia.org/wiki/Confidence_interval)For α = 0.05One cannot say: "with probability .95 the parameter μ lies in the confidence interval." One only knows that by repetition in 95 % of the cases, μ will be in the calculated interval. In 5% of the cases however it does not. And unfortunately one does not know in which of the cases this happens. That is why one can say: "with confidence level 95%, μ lies in the confidence interval."

How to Use Confidence Intervals



The Confidence Interval and Statistical Significance: retrieved from http://sportsci.org/resource/stats/index.html

Confidence Interval for Two Independent Samples, Continuous Outcomehttp://sphweb.bumc.bu.edu/otlt/MPH-Modules/BS/BS704_Confidence_Intervals/BS704_Confidence_Intervals5.html

Confidence Intervals for Matched Samples, Continuous Outcomehttp://sphweb.bumc.bu.edu/otlt/MPH-Modules/BS/BS704_Confidence_Intervals/BS704_Confidence_Intervals6.html


http://sphweb.bumc.bu.edu/otlt/MPH-Modules/BS/BS704_Confidence_Intervals/BS704_Confidence_Intervals6.html




http://sportsci.org/resource/stats/index.html

Unit 6: Appendix: Effect Size for repeated measures versus two-group T-tests

Source: http://web.uccs.edu/lbecker/Psy590/es.htmEffect Size (ES)

I. Overview II. Effect Size Measures for Two Independent Groups1. Standardized difference between two groups. 2. Correlation measures of effect size. 3. Computational examples III. Effect Size Measures for Two Dependent Groups.IV. Meta AnalysisV. Effect Size Measures in Analysis of VarianceVI. ReferencesEffect Size CalculatorsAnswers to the Effect Size Computation Questions

I. Overview

Effect size (ES) is a name given to a family of indices that measure the magnitude of a treatment effect. Unlike significance tests, these indices are independent of sample size. ES measures are the common currency of meta-analysis studies that summarize the findings from a specific area of research. See, for example, the influential meta-analysis of psychological, educational, and behavioral treatments by Lipsey and Wilson (1993).

There is a wide array of formulas used to measure ES. For the occasional reader of meta-analysis studies, like myself, this diversity can be confusing. One of my objectives in putting together this set of lecture notes was to organize and summarize the various measures of ES.

In general, ES can be measured in two ways:

a) as the standardized difference between two means, or

b) as the correlation between the independent variable classification and the individual scores on the dependent variable. This correlation is called the "effect size correlation" (Rosnow & Rosenthal, 1996).

These notes begin with the presentation of the basic ES measures for studies with two independent groups. The issues involved when assessing ES for two dependent groups are then described.

II. Effect Size Measures for Two Independent Groups1. Standardized difference between two groups.

Cohen's dd = M1 - M2 /

where

= [(X - M)² / N]

where X is the raw score,

M is the mean, and N is the number of

cases.

Cohen (1988) defined d as the difference between the means, M1 - M2, divided by standard deviation, , of either group. Cohen argued that the standard deviation of either group could be used when the variances of the two groups are homogeneous.

In meta-analysis the two groups are considered to be the experimental and control groups. By convention the subtraction, M1 - M2, is done so that the difference is positive if it is in the direction of improvement or in the predicted direction and negative if in the direction of deterioration or opposite to the predicted direction.

d is a descriptive measure.

d = M1 - M2 / pooled pooled = [(1²+ ²) / 2]

In practice, the pooled standard deviation, pooled, is commonly used (Rosnow and Rosenthal, 1996).

The pooled standard deviation is found as the root mean square of the two standard deviations (Cohen, 1988, p. 44). That is, the pooled standard deviation is the square root of the average of the squared standard deviations. When the two standard deviations are similar the root mean square will be not differ much from the simple


http://web.uccs.edu/lbecker/Psy590/es.htm#Cohen%23Cohen

http://web.uccs.edu/lbecker/Psy590/es.htm#Rosnow%20&%20Rosenthal%23Rosnow%20&%20Rosenthal




http://web.uccs.edu/lbecker/Psy590/es.htm#Lipsey%20&%20Wilson%23Lipsey%20&%20Wilson


average of the two variances.

d = 2t(df) or

d = tn1 + n2) / (df)(n1n2)]

d can also be computed from the value of the t test of the differences between the two groups (Rosenthal and Rosnow, 1991). . In the equation to the left "df" is the degrees of freedom for the t test. The "n's" are the number of cases for each group. The formula without the n's should be used when the n's are equal. The formula with separate n's should be used when the n's are not equal.

d = 2r / (1 - r²) d can be computed from r, the ES correlation.

d = g(N/df) d can be computed from Hedges's g.The interpretation of Cohen's d

Cohen's Standard

Effect Size

Percentile Standing

Percent of Nonoverla

p 2.0 97.7 81.1% 1.9 97.1 79.4% 1.8 96.4 77.4% 1.7 95.5 75.4% 1.6 94.5 73.1% 1.5 93.3 70.7% 1.4 91.9 68.1% 1.3 90 65.3% 1.2 88 62.2% 1.1 86 58.9% 1.0 84 55.4% 0.9 82 51.6%

LARGE 0.8 79 47.4% 0.7 76 43.0% 0.6 73 38.2%

MEDIUM 0.5 69 33.0% 0.4 66 27.4% 0.3 62 21.3%

SMALL 0.2 58 14.7% 0.1 54 7.7% 0.0 50 0%

Cohen (1988) hesitantly defined effect sizes as "small, d = .2," "medium, d = .5," and "large, d = .8", stating that "there is a certain risk in inherent in offering conventional operational definitions for those terms for use in power analysis in as diverse a field of inquiry as behavioral science" (p. 25).

Effect sizes can also be thought of as the average percentile standing of the average treated (or experimental) participant relative to the average untreated (or control) participant. An ES of 0.0 indicates that the mean of the treated group is at the 50th percentile of the untreated group. An ES of 0.8 indicates that the mean of the treated group is at the 79th percentile of the untreated group. An effect size of 1.7 indicates that the mean of the treated group is at the 95.5 percentile of the untreated group.

Effect sizes can also be interpreted in terms of the percent of nonoverlap of the treated group's scores with those of the untreated group, see Cohen (1988, pp. 21-23) for descriptions of additional measures of nonoverlap.. An ES of 0.0 indicates that the distribution of scores for the treated group overlaps completely with the distribution of scores for the untreated group, there is 0% of nonoverlap. An ES of 0.8 indicates a nonoverlap of 47.4% in the two distributions. An ES of 1.7 indicates a nonoverlap of 75.4% in the two distributions.

3. Computational Examples

The following data come from Wilson, Becker, and Tinker (1995). In that study participants were randomly assigned to either EMDR treatment or delayed EMDR treatment. Treatment group assignment is called TREATGRP in the analysis below. The dependent measure is the Global Severity Index (GSI) of the Symptom Check List-90R. This index is called GLOBAL4 in the analysis below. The analysis looks at the the GSI scores immediately post treatment for those assigned to the EMDR treatment group and at the second pretreatment testing for those assigned to the


http://web.uccs.edu/lbecker/Psy590/es.htm#Wilson%23Wilson



http://web.uccs.edu/lbecker/Psy590/es.htm#Rosenthal%20&%20Rosnow%23Rosenthal%20&%20Rosnow


delayed treatment condition. The output from the SPSS MANOVA and CORR(elation) procedures are shown below. Cell Means and Standard Deviations Variable .. GLOBAL4 GLOBAL INDEX:SLC-90R POST-TEST FACTOR CODE Mean Std. Dev. N 95 percent Conf. Interval TREATGRP TREATMEN .589 .645 40 .383 .795 TREATGRP DELAYED 1.004 .628 40 .803 1.205 For entire sample .797 .666 80 .648 .945* * * * * * * * * * * * * A n a l y s i s o f V a r i a n c e -- Design 1 * * * * * * * * * * * * Tests of Significance for GLOBAL4 using UNIQUE sums of squares Source of Variation SS DF MS F Sig of F WITHIN CELLS 31.60 78 .41 TREATGRP 3.44 1 3.44 8.49 .005 (Model) 3.44 1 3.44 8.49 .005 (Total) 35.04 79 .44 - - Correlation Coefficients - -GLOBAL4TREATGRP .3134 ( 80) P= .005

Look back over the formulas for computing the various ES estimates. This SPSS output has the following relevant information: cell means, standard deviations, and ns, the overall N, and MSwithin. Let's use that information to compute ES estimates.

d = M1 - M2 / [( 1² +²)/ 2] = 1.004 - 0.589 / [(0.628² + 0.645²) / 2] = 0.415 / [(0.3944 + 0.4160) / 2] = 0.415 / (0.8144 / 2) = 0.415 / 0.4052 = 0.415 / = .65

Cohen's d Cohen's d can be computed using the two standard deviations.

What is the magnitude of d, according to Cohen's standards?

The mean of the treatment group is at the _____ percentile of the control group.

III. Effect Size Measures for Two Dependent Groups.

There is some controversy about how to compute effect sizes when the two groups are dependent, e.g., when you have matched groups or repeated measures. These designs are also called correlated designs. Let's look at a typical repeated measures design.

A Correlated (or Repeated Measures) Design

OC1 OC2

OE1 X OE2

Participants are randomly assigned to one of two conditions, experimental (E.) or control (C.). A pretest is given to all

participants at time 1 (O.1.). The treatment is administered at "X". Measurement at time 2 (OE2) is posttreatment for the

experimental group. The control group is measured a second time at (OC2) without an intervening treatment.. The time period between O.1 and O.2 is the same for both groups.

This research design can be analyzed in a number of ways including by gain scores, a 2 x 2 ANOVA with measurement time as a repeated measure, or by an ANCOVA using the pretest scores as the covariate. All three of these analyses make use of the fact that the pretest scores are correlated with the posttest scores, thus making the significance tests more sensitive to any differences that might occur (relative to an analysis that did not make use of the correlation between the pretest and posttest scores).



An effect size analysis compares the mean of the experimental group with the mean of the control group. The experimental group mean will be the posttreatment scores, OE2. But any of the other three means might be used as the control group mean. You could look at the ES by comparing OE2 with its own pretreatment score, OE1, with the pretreatment score of the control group, OC1, or with the second testing of the untreated control group, OC2. Wilson, Becker, and Tinker (1995) computed effect size estimates, Cohen's d, by comparing the experimental group's posttest scores (OE2) with the second testing of the untreated control group (OC2). We choose OC2 because measures taken at the same time would be less likely to be subject to history artifacts, and because any regression to the mean from time 1 to time 2 would tend to make that test more conservative.

Suppose that you decide to compute Cohen's d by comparing the experimental group's pretest scores (OE2) with their own pretest scores (OE1), how should the pooled standard deviation be computed? There are two possibilities, you might use the original standard deviations for the two means , or you might use the paired t-test value to compute Cohen's d. Because the paired t-test value takes into account the correlation between the two scores the paired t-test will be larger than a between groups t-test. Thus, the ES computed using the paired t-test value will always be larger than the ES computed using a between groups t-test value, or the original standard deviations of the scores. Rosenthal (1991) recommended using the paired t-test value in computing the ES. A set of meta-analysis computer programs by Mullen and Rosenthal (1985) use the paired t-test value in its computations. However, Dunlop, Cortina, Vaslow, & Burke (1996) (reference inserted below) convincingly argue that the original standard deviations (or the between group t-test value) should be used to compute ES for correlated designs. They argue that if the pooled standard deviation is corrected for the amount of correlation between the measures, then the ES estimate will be an overestimate of the actual ES. As shown in Table 2 of Dunlop et al., the overestimate is dependent upon the magnitude of the correlation between between the two scores. For example, when the correlation between the scores is at least .8, then the ES estimate is more than twice the magnitude of the ES computed using the original standard deviations of the measures.

The same problem occurs if you use a one-degree of freedom F value that is based on a repeated measures to compute an ES value.

In summary, when you have correlated designs you should use the original standard deviations to compute the ES rather than the paired t-test value or the within subject's F value.

Dunlop, W. P., Cortina, J. M., Vaslow, J. B., & Burke, M. J. (1996). Meta-analysis of experiments with matched groups or repeated measures designs. Psychological Methods, 1, 170-177.


http://web.uccs.edu/lbecker/Psy590/es.htm#Dunlop%20et.al.%23Dunlop%20et.al.

http://web.uccs.edu/lbecker/Psy590/es.htm#Dunlop%20et.al.%23Dunlop%20et.al.

http://web.uccs.edu/lbecker/Psy590/es.htm#Mullen%20&%20Rosenthal%23Mullen%20&%20Rosenthal

http://web.uccs.edu/lbecker/Psy590/es.htm#Rosenthal%23Rosenthal

http://web.uccs.edu/lbecker/Psy590/es.htm#Rosenthal%23Rosenthal

http://web.uccs.edu/lbecker/Psy590/es.htm#Wilson%23Wilson

Unit 7: One-Way ANOVA

WHAT IF YOU HAVE MORE THAN TWO GROUPS TO COMPARE?

Group 1 Group 2 Group 3Normal diet (control

group)Normal diet + typical

energy barNormal diet + typical energy bar with high

Mg++

Performance with no special psychological instruction (control group)

Performance with self help individual psychological guide worksheet activities

Performance with GROUP psychological guide worksheet activities

Analysis of Variance (ANOVA)

s2 = variance

If the variance (variability) between the groups is large, compared to the variance (variability) within the groups, then there may be a significant difference between the groups

Step 1: Ask your question12 people participate in a stress management program, each is randomly assigned to one of three treatments

1) no treatment (control)2) self study and help3) private discussion with counselor

Is there a difference in stress level in a self report stress assessment? (low score is lower stress)

Null hypothesis : HO the means of the groups are equal1 = 2 = 3

Step 2: Collect your data

no treatment (control)

self study and help

private discussion with

counselor5 2 93 5 74 2 84 3 8

MEAN 4 3 8

This (above example) is a “Between Groups” ANOVA .



We seldom (if ever?) use this design of experiment, and statistical analysis, in experimental research because there is no effective control group (see unit 11). Accordingly, the SPSS analysis of this design will not be discussed.

If you need to do a Between ANOVA in SPSS use:Analyse => GLM => univariate. Select the options of “descriptives, effect size, homogeneity test”, and obtain post-hoc test through the “Options” button, select the independent variable from the box labeled “estimated Marginal Means:Factors(s) and Factor interactions” and transfer it to the box labeled “Display means for”. Once variable is transferred select the “Compare main effects” box. Then in the “confidence interval adjustment” and select the Bonferonni correction (Source: Discovering Statistics using SPSS for Windows, (2000) Andy Field, Sage, London., p298-299)

TERMINOLOGY: 1 way ANOVAa single independent variable = FACTOR = main effect

Multiple (>2) groups (1 independent variable = 1 FACTOR = 1 main effect of this factor)

none self counselorAnswers to questions you wanted to ask about testing groups for comparisons of means (but were afraid to ask):

Q: can you do an ANOVA on two groups?

YES, in this case it is doing the same thing as a T-test

Q: If you have 3 groups, can you do a T-test on each pair combination to look for sig differences between them (skip the ANOVA stuff)?NO, repeated application of t-test lead to increased chances of error due to probability pyramiding.

Another one-way ANOVA example study:REPEATED MEASURES ANOVA DESIGN(within subjects design)

Each of 3 different diets tested in each subject:Diet 1 Diet 2 Diet 3

Normal diet (control group)

Normal diet + super power builder

Normal diet + super power builder with high magnesium



Use REPEATED MEASURES ANOVA (for within subjects design)Ask your question (This is the same example as practiced in SPSS lesson)Is there a difference in ventilation rate during the 10th minute of steady state exercise at a 2 mph on a treadmill, for subjects with a control/normal diet, normal diet + pre run energy bar, & normal diet + pre run pizza?

Collect your dataSubject T1VE T2VE T3VE

1 51.68 57.22 53.902 58.03 60.23 61.593 47.01 51.06 52.544 56.98 56.71 55.935 63.06 61.87 58.816 47.02 52.76 50.327 46.45 49.55 52.968 50.81 45.76 46.019 50.57 51.03 51.9010 38.07 44.14 44.8511 50.22 45.38 52.2512 67.93 60.60 60.9813 49.15 51.27 48.7414 44.43 41.99 41.5615 50.47 55.50 60.1416 61.35 46.61 47.8517 59.51 56.55 53.8018 43.83 41.00 43.7719 61.34 59.00 53.7420 60.02 61.28 59.5621 60.39 58.92 51.73

Run statistical analysis using SPSSThese steps to run repeated measures ANOVA with SPSS will be practiced in the SPSS lesson in computer lab.

assign labels to the T1VE, T2VE, T3VE variables to make output easier to read

t1ve = Minute ventilation – controlt2ve = Minute ventilation – energy bart3ve = Minute ventilation - pizza

analyze >> general linear model >> repeated measureswithin subject factor name – PREDIET (for this example)number of levels - 3 (in this example)adddefineselect t1ve & click on top right direction arrow (to put test one

into first place for within subject variables)repeat for t2ve & t2veOptions >> select descriptive statistics, estimates of effect size,

& Observed PowercontinuePlots >> PREDIET > onto horizontal axis >> Add



continueOK (to run the test)

Interpret SPSS output to find answer to your question (interpretation of output is added within the output using ITALICS).

General Linear ModelWithin-Subjects Factors

Measure: MEASURE_1

t1vet2vet3ve

prediet123

DependentVariable

Descriptive Statistics

53.2533 7.70010 21

52.7824 6.69095 21

52.5205 5.73494 21

Minute Ventilation at 10thminute, CONTROL DIETMinute Ventilation at 10thminute, PRE RUNENERGY BAR DIETMinute Ventilation at 10thminute, PRE RUN PIZZADIET

Mean Std. Deviation N

REVIEW YOUR GROUP MEANSMultivariate Testsb

.017 .160a 2.000 19.000 .853 .017

.983 .160a 2.000 19.000 .853 .017

.017 .160a 2.000 19.000 .853 .017

.017 .160a 2.000 19.000 .853 .017

Pillai's TraceWilks' LambdaHotelling's TraceRoy's Largest Root

Effectprediet

Value F Hypothesis df Error df Sig.Partial EtaSquared

Exact statistica.

Design: Intercept Within Subjects Design: prediet

b.

IGNORE THE ABOVE TABLE



Mauchly's Test of Sphericityb

Measure: MEASURE_1

.666 7.729 2 .021 .749 .797 .500Within Subjects Effectprediet

Mauchly's WApprox.

Chi-Square df Sig.Greenhouse-Geisser Huynh-Feldt Lower-bound

Epsilona

Tests the null hypothesis that the error covariance matrix of the orthonormalized transformed dependent variables isproportional to an identity matrix.

May be used to adjust the degrees of freedom for the averaged tests of significance. Corrected tests are displayed inthe Tests of Within-Subjects Effects table.

a.

Design: Intercept Within Subjects Design: prediet

b.

Look at Greenhouse-Geisser value in above table. If <1 then use G-G line in next table, if G-G = 1 then use Sphericity assumed line in next table

Tests of Within-Subjects Effects

Measure: MEASURE_1

5.792 2 2.896 .247 .783 .0125.792 1.499 3.864 .247 .719 .0125.792 1.593 3.635 .247 .733 .0125.792 1.000 5.792 .247 .625 .012

469.792 40 11.745469.792 29.980 15.670469.792 31.867 14.742469.792 20.000 23.490

Sphericity AssumedGreenhouse-GeisserHuynh-FeldtLower-boundSphericity AssumedGreenhouse-GeisserHuynh-FeldtLower-bound

Sourceprediet

Error(prediet)

Type III Sumof Squares df Mean Square F Sig.

Partial EtaSquared

Look for significance in the "Within subjects" table FOR THE DEPENDENT VARIABLE OF INTEREST (PREDIET IN THIS CASE), using the appropriate line (G-G or Sphericity assumed, as explained above).

Tests of Within-Subjects Contrasts

Measure: MEASURE_1

5.639 1 5.639 .331 .572 .016.153 1 .153 .024 .879 .001

341.139 20 17.057128.653 20 6.433

predietLinearQuadraticLinearQuadratic

Sourceprediet

Error(prediet)


Partial EtaSquared




Tests of Between-Subjects Effects

Measure: MEASURE_1Transformed Variable: Average

175980.459 1 175980.459 1551.032 .000 .9872269.205 20 113.460

SourceInterceptError


Partial EtaSquared


Profile Plots

1 2 3

prediet

52.5

52.6

52.7

52.8

52.9

53.0

53.1

53.2

53.3

Estim

ated

Mar

gina

l Mea

ns

Estimated Marginal Means of MEASURE_1

THIS GRAPH OF THE GROUP MEANS ILLUSTRATES YOUR RESULTS, IT WILL BE MORE CRITICAL IN THE NEXT UNIT EXAMINING 2-WAY ANOVAS.

How to do post hoc tests for the one-way repeated measures ANOVA

Method 1) Do paired t tests between each of the measures, using Bonferonni correction (i.e. level of significance = 0.05/# of comparisons) (references: How to Use SPSS: A Step-by-Step Guide to Analysis and Interpretation (Third Edition) Brian C. Cronk, ISBN 1-884585-55-8; © 2004, Pyrczak Publishing, pg 72; Winter, E. M., Eston, R. G., & Lamb, K. L. (2001). Statistical analyses in the physiology of exercise and kinanthropometry. J Sports Sci, 19(10), 761-75. (latter is an excellent discussion of repeated measures ANOVA, post hoc testing: see pg 771-773))

Method 2) See pg 180 Stats in Kinesiology 2nd ed.

Method 3) For SPSS V 7.5 and later. Obtain post-hoc test through the “Options” button, select the repeated

measures independent variable from the box labeled “Estimated



Marginal Means:Factors(s) and Factor interactions” and transfer it to the box labeled “Display means for”. Once variable is transferred select the “Compare main effects” box. Then in the “confidence interval adjustment” select the Bonferonni correction (Source: Discovering Statistics using SPSS for Windows, (2000) Andy Field, Sage, London., p 331)

Opinions vary widely on how to do post-hoc testing in a repeated measures design, consult with you thesis chair for advice on the method to use.

Alternately: do not do post-hoc tests, do planned comparisons (more powerful if well planned because you only do the needed comparisons).(See: Discovering Statistics using SPSS for Windows, (2000) Andy Field, Sage, London., p 328 - 329). If you select the “repeated” contrast you will compare each level against the previous level.

WHAT ABOUT THE MAGNITUDE OF THE TREATMENT EFFECT? (ANOVA version of the T-test Cohen’s D, or “effect size”)

From above example:Tests of Within-Subjects Effects

Measure: MEASURE_1

5.792 2 2.896 .247 .783 .0125.792 1.499 3.864 .247 .719 .0125.792 1.593 3.635 .247 .733 .0125.792 1.000 5.792 .247 .625 .012

469.792 40 11.745469.792 29.980 15.670469.792 31.867 14.742469.792 20.000 23.490

Sphericity AssumedGreenhouse-GeisserHuynh-FeldtLower-boundSphericity AssumedGreenhouse-GeisserHuynh-FeldtLower-bound

Sourceprediet

Error(prediet)


Partial EtaSquared

Look for significance in the "Within subjects" table, using the appropriate line (G-G or Sphericity assumed, as explained above), for the prediet row (i.e. your groups).

Look at the “Partial Eta Squared” (in addition to the “Sig”)

Interpretation of Magnitude of Treatment Effect in an ANOVA using Partial Eta Squared

1. Vincent 1999 (Statistics in Kinesiology, using Cohen standards for the behavioral sciences), Report Eta Squared value in the Within-Subjects effects table. This number approximates (overestimates somewhat) the Omega Squared value for the ANOVA



2 > .15 is large 2 > .06 is medium 2 > .01 is small

2. Maher, J. M., Markey, J. C., & Ebert-May, D. (2013). The Other Half of the Story: Effect Size Analysis in Quantitative Research. CBE Life Sciences Education, 12(3), 345–351. http://doi.org/10.1187/cbe.13-04-0082

Table 2.

Interpreting effect size valuesa

Effect size measure Small effect size

Medium effect size

Large effect size

Very large effect size

Odds ratio 1.5 2.5 4 10Cohen's d (or one of its variants) 0.20 0.50 0.80 1.30

r 0.10 0.30 0.50 0.70Cohen's f 0.10 0.25 0.40 —Eta-squared 0.01 0.06 0.14 —a,Cohen, 1992 , 1988 ; Rosenthal, 1996 .

3. Ferguson, Christopher J. An effect size primer: A guide for clinicians and researchers. Professional Psychology: Research and Practice, Vol 40(5), Oct 2009, 532-538. http://dx.doi.org/10.1037/a0015808

NOTE: Also use the Eta-squared value to assess magnitude of treat effect in more complex (2-way) ANOVAs (discussed subsequently).


http://psycnet.apa.org/doi/10.1037/a0015808

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3763001/#B29



http://doi.org/10.1187/cbe.13-04-0082

Unit 8: Two Way ANOVA

TERMINOLOGY: 1 way ANOVA (in this example ANOVA done on 2 groups)

a single independent variable = FACTOR

Question 1: What is the effect of the supplement (IV = independent variable) on HRrest (DV= dependent variable)?

INDEPENDENT VAR= Effect of treatment

IV Control Zinc Supplements

DV HRrest53 ±16

HRrest55 ±13

Question 2: What is the effect of gender (IV) on HRrest (DV)?IV DV

INDEPENDENT VAR

= Effect of GENDER

Males (n=10) HRrest53 ±16

Females (n=8) HRrest57 ±18

Add 2, 1-way ANOVAs together to get a 2-way ANOVA

FACTORIAL ANOVA = 2 way

two independent variables = 2 FACTORS = 2 main effects

INDEPENDENT VAR 1= factor 1= Effect of supplement = Main effect of supplementControl Zinc

SupplementsINDEPENDENT

VAR 2= factor 2 = Effect of GROUP = main effect of group

Males (n=10) HRrest53 ±16

HRrest55 ±13

Females (n=8)HRrest57 ±18

HRrest45 ±27

Question 1: What is the (main) effect of the supplement on the HRrest?

Question 2: What is the (main) effect of gender on the HRrest?Question 3: Is there an INTERACTION between gender and

supplementation?



****= do both of the groups respond similarly to the treatment (of supplements, of training, of mental practice etc.) *********

In a factorial ANOVA: Always examine the interaction first. If it is sig, then do not look at main effects (or at least the interpretation of the main effects must be qualified) (Thomas & Nelson, 4th ed, pg 154) (Using SPSS for Windows, S Green, N Salkind, T Akey, 1997, Prentice Hall, pgs 209)

Example #2 from MANY thesis projectsMIXED FACTORIAL ANOVA = 2 way between-within ANOVA (2 groups, each tested twice)two independent variables = 2 FACTORS = 2 main effectsone of the independent variables is a repeated measures

INDEPENDENT VAR 1= factor 1= Main Effect of time of experiment – REPEATED (within) MEASUREPre Measure Post Measure

INDEPENDENT VAR 2

= factor 2 =Main Effect of GROUPBETWEEN MEASURE

Control (n=10)HRrest53 ±16

orSelf Confidence

HRrest55 ±13

orSelf Confidence

Training (n=8)(eg. mental practice or physical exercise)

HRrest57 ±18

orSelf Confidence

HRrest45 ±27

orSelf Confidence

Question 3: Is there an INTERACTION between group and time over exp?

☺☺☺☺= do both of the groups respond similarly during the time of the experiment?☺☺☺☺☺☺☺

Null Hypothesis should be worded in the following way:

There is no difference in the change in dependent variable(s) over the course of the experiment, when comparing the control and treatment groups.

OR similar such as:



When comparing the control and treatment groups, there is no difference over the course of the experiment in the change in dependent variable(s).

Question 1: What is the effect of the time of the experiment (pre to post time) on the HRrest (or on self confidence) – FOR SUBJECTS FROM BOTH GROUPS COMBINED


Control + Training (n=18)

HRrest55 ±17

orSelf Confidence

HRrest50 ±20

orSelf Confidence

Question 2: What is the effect of group on the HRrest – FOR BOTH PRE AND POST MEASURES COMBINED (=is there a difference in HRrest/self confidence between the groups?)

Pre & Post Measures

INDEPENDENT VAR 2


Control (n=10)HRrest

54 ±14.5 or

Self ConfidenceTraining (n=8)(eg. mental practice or physical exercise)

HRrest51 ±23

orSelf Confidence



Example #1: two-way ANOVA, between-between. (Shenelle Cardiac rehab data)Ask your questionShenelle (Graduate student 2003) asked if there was a difference in level of exercise performed by people following their cardiac event (e.g., heart attack) if they participated in the St. Joe’s cardiac rehabilitation program, versus if they did not participate in the program. She surveyed people after their heart attacks at three different time points. For this course we will break her data into two parts, to produce 2, 2x2 ANOVAs (one for this SPSS lesson, and one for the related SPSS assignment), as opposed to analyzing it as a 2x3 ANOVA as she did.

For this example (This is the same example as practiced in SPSS lesson unit 8 example #1) the subjects at time 1 (= first measure at 5-7 months post cardiac event) and time 2 (= second measure at 11-13 months post cardiac event) are examined. To get the data at the two time points she did not survey the same people (this would have taken two years for data collection), she surveyed (in the same month) different people who were in the appropriate time frames, post heart attack. Hence, this is a between-between ANOVA.

BETWEEN MEASURES factor 1Main EffectTime of measurement(one categorical Column in SPSS)

time 1 (= first measure at 5-7 months post cardiac event)

time 2 (= second measure at 11-13 months post cardiac event)

BETWEEN MEASURES factor 2Main EffectParticipartion in program(one categorical Column in SPSS)

People who participated in the cardiac rehabilitation programPeople who did not participated in the cardiac rehabilitation program



Collect your dataGroup Time Dependent variable1=cardiac rehab program participation

1= first measure at 5-7 months post cardiac event

Kcals expended per week in moderate or greater intensity exercise

2=no cardiac rehab program participation

2= second measure at 11-13 months post cardiac event

Group Time kcals1 1 1671.71 1 593.91 1 1270.51 1 1653.61 1 814.51 1 1890.91 1 1658.21 1 218.21 1 1216.41 1 703.61 2 283.61 2 3112.91 2 557.61 2 909.11 2 840.01 2 2013.21 2 965.21 2 773.61 2 970.51 2 553.62 1 127.32 1 0.02 1 0.02 1 0.02 1 872.72 1 0.02 1 427.32 1 982.52 1 0.02 1 200.52 2 745.52 2 474.52 2 265.52 2 581.82 2 0.02 2 189.12 2 0.02 2 420.02 2 0.02 2 0.0



Run statistical analysis using SPSSThese steps to run repeated measures ANOVA with SPSS will be practiced in the SPSS lesson in computer lab.When you have a “between” ANOVA, you use grouping variables to establish what group a measure belongs in. The grouping variable must be numeric (e.g., 0, 1, 2), SPSS won't allow character grouping variable (e.g, m/f). You then make the output easily meaningful by setting “value” labels to the grouping variable values.

For the GROUP variable, set the VALUES as:1=cardiac rehab program participation2=no cardiac rehab program participation

For the TIME variable, set the VALUES as:1= first measure at 5-7 months post cardiac event2= second measure at 11-13 months post cardiac event

Set “label” for dependent variable (Energy expended in moderate of more intense exercise)

Analyze > GENERAL LINEAR MODEL > UnivariateSelect dependent variable (kcals) and place in Dependent boxSelect first independent variable (group) and place in Fixed Factor

boxSelect second independent variable (time) and place in Fixed Factor

boxSelect Plots

Select & enter factor for horizontal axis (time in this example)Select & enter factor for separate lines (group in this example)Click AddContinue

Select OptionsCheck, Display Descriptive Statistics & Display Estimate of

effect sizeIf you have >2 levels in one of your factors, and are going to do

post-hoc tests in SPSS, then you need to run the Levene’s test of homogeneity of variance, this is NOT included in the following sample printout (Some supervisors may ask for this test to be run even if you have only 2 levels in your factors).

continueClick OK (to run analysis)



Interpret SPSS output to find answer to your question.Univariate Analysis of Variance

Between-Subjects Factors

cardiacrehabprogramparticipation

20

no cardiacrehabprogramparticipation

20

firstmeasure at5-7 monthspostcardiacevent

20

secondmeasure at11-13monthspostcardiacevent

20

1.00

2.00

group

1.00

2.00

time

Value Label N

ABOVE TABLE NOT CRITICALDescriptive Statistics

Dependent Variable: Energy expended in moderate of more intense exercise

1169.1500 561.50369 10

1097.9300 842.64389 10

1133.5400 697.86809 20

261.0300 377.79121 10

267.6400 276.03323 10

264.3350 322.04146 20

715.0900 658.76771 20

682.7850 744.20905 20

698.9375 693.91225 40

timefirst measure at5-7 months postcardiac eventsecond measureat 11-13 monthspost cardiac eventTotalfirst measure at5-7 months postcardiac eventsecond measureat 11-13 monthspost cardiac eventTotalfirst measure at5-7 months postcardiac eventsecond measureat 11-13 monthspost cardiac eventTotal

groupcardiac rehabprogram participation

no cardiac rehabprogram participation

Total


ABOVE TABLE : RECORD YOUR MEANS AND s VALUES




Dependent Variable: Energy expended in moderate of more intense exercise

7580753.223a 3 2526917.741 8.123 .000 .40419540545.2 1 19540545.16 62.818 .000 .636

7555173.320 1 7555173.320 24.288 .000 .40310436.130 1 10436.130 .034 .856 .00115143.772 1 15143.772 .049 .827 .001

11198301.0 36 311063.91538319599.3 4018779054.2 39

SourceCorrected ModelInterceptgrouptimegroup * timeErrorTotalCorrected Total


Partial EtaSquared

R Squared = .404 (Adjusted R Squared = .354)a.

ABOVE TABLE LOOK AT (IN THIS ORDER)1. THE INTERACTION OF THE MAIN FACTORS (GROUP & TIME) IF the interaction is not sig, then proceed with the next 2. The main effect of the first factor (group) 3. The main effect of the second factor (time)

In this example #1: interaction is not sig, main effect of time is not sig, main effect of group is sig.



Profile Plots

first measure at 5-7 months post cardiac event

second measure at 11-13 months post cardiac event

time

200.00

400.00

600.00

800.00

1000.00

1200.00

Estim

ated

Mar

gina

l Mea

ns

groupcardiac rehab program participationno cardiac rehab program participation

Estimated Marginal Means of Energy expended in moderate of more intense exercise

ABOVE PLOT : Important illustration of your results. If the lines are parallel (or nearly parallel) there is no interaction between the two factors. If the lines are not parallel (or nearly parallel) there is an interaction.

How to do post hoc tests for the two-way between-between ANOVA?

You only need to do post-hoc tests if you have more than 2 levels in one of your factors (Source: Discovering Statistics using SPSS for Windows, (2000) Andy Field, Sage, London., p 314-315).

Methods for post hoc testing following a sig interaction are not agreed upon by researchers (Thomas & Nelson, 4th ed, pg 154)

Method 1) Describe the profile plot and the interaction (Thomas & Nelson, 4th ed, pg 153)

Method 2) Examine simple main effects using the Scheffé procedure (Thomas & Nelson, 4th ed, pg 153)

Method 3) Use a tetrad contrast for four cell means (if you have a 2x2 ANOVA) (Using SPSS for Windows, S Green, N Salkind, T Akey, 1997, Prentice Hall, pgs 219-222)

Method 4) Do post hoc for main effects with > 2 levels using SPSS. (Source: Discovering Statistics using SPSS for Windows, (2000) Andy Field, Sage,



London., p 314-315). Use “post-hoc” button => select one post hoc for equal var assumed based on the following:

Bonferonni, if # of comparisons is small Tukey, if # comparisons is large REGWQ, best unless group sizes are different.

=> select one post hoc for equal var not assumed: Games-Howell (liberal if sample sizes are small, good if sample

sizes are unequal). Tamhanes T2 (conservative)

You will have selected the option of the homogeneity of variance test, so you will know which post hoc test category to pay attention to.

If you have no sig interaction, and a sig main effect with more than 2 levels for the main effect: Follow up the sig main effect with a Scheffé test (Thomas & Nelson, 4th

ed, pg 153)

Discuss your post hoc testing procedures with your thesis advisor.

Example #2: two-way ANOVA, between-within. (April, mental practice data)Ask your questionApril (Graduate student 2002) asked if there was a difference in distance rowed on an ergometer in 5 minutes in an ergometer rowing competition, if an athlete was trained in using a precompetition visualization routine. All subjects were tested for maximum rowing distance in 5 minutes (pre test). Subjects were then randomly assigned to the group that received the visualization training, or did not receive the training. After 12 weeks, during which the visualization group (group 1) did visualization training plus regular rowing team training, while the no visualization group (group 2) did regular rowing team training, all subjects were tested again (post test). This is the same example as practiced in SPSS lesson unit 8 example #2.

Collect your data1=precompetition visualization routine practiced2=no precompetition visualization routine practiced

Group Subj #

Average Distance Rowed (Meters) Pre Test

Average Distance Rowed (Meters) Post Test

1 1 1512.5 1561.51 7 1454 1503.51 8 1364.5 1432.51 11 1412.5 1496.51 12 1494.5 1604.51 15 1469.5 15321 16 1431 1463.51 18 1691.5 1754.51 21 1815.5 1860.51 24 1885.5 1926.5



2 4 1420 14672 5 1311 14082 6 1322.5 13962 9 1482 1522.52 10 1424.5 1514.52 13 1485.5 15232 17 1747.5 1805.52 20 1710.5 1739.52 23 1717.5 17842 25 1653.5 1710.52 26 1768.5 1839.5

Run statistical analysis using SPSSThese steps to run repeated measures ANOVA with SPSS will be practiced in the SPSS lesson in computer lab.When you have a “between” factor in an ANOVA, you use grouping variables to establish what group a measure belongs in. The grouping variable must be numeric (e.g., 0, 1, 2), SPSS won't allow character grouping variable (e.g, m/f). You then make the output easily meaningful by setting “value” labels to the grouping variable values.

For the GROUP variable, set the VALUES as:1=precompetition visualization routine practiced2=no precompetition visualization routine practiced

When you have a “within” factor in an ANOVA, you use a separate columns for each within measure of your dependant variable.

Set “label” for dependent variables, to make output easy to read.

Analyze > GENERAL LINEAR MODEL > REPEATED MEASURESWithin-subject factor name:

test This is a name you give for organizational purposes for the WITHIN effect, use “test” for pre & post tests for this example

Number of levels 2 This is the # of WITHIN levels, 2 if only pre & post, 3 if pre, mid, post etc.

Hit the ADD buttonYou enter name for the dependent variable (“Measure Name” box), but don’t use the

exact name used in the columns in the data file. The name here is for organization and selection. In this example use “distance”. Enter one measure name and click ADD.

Click DEFINE buttonIn this new window:



Move the two within subject tests into the “Within-Subjects Variables” box, in the correct order (i.e., pre goes first, post goes second). In this example it appears as follows:pre(1,distance)post(2, distance)

Move the one grouping variable into the “Between-Subjects Factor” box. In this example it appears as follows:Group[group]

Hit the PLOTS buttonPut within factor (“test” in this example) as horizontal axis. Put between grouping variable name (“group” in this example) as

separate lines.Click ADDClick CONTINUEclick OPTIONSCheck, Display Descriptive Statistics & Display Estimate of

effect sizeClick CONTINUE

Click OK (to run analysis)

Interpret SPSS output to find answer to your question.

General Linear ModelWithin-Subjects Factors

Measure: distance

prepost

test12

DependentVariable

ABOVE TABLE NOT CRITICALBetween-Subjects Factors

visualization routine 10

NOvisualization routine

11

1.00

2.00

GroupValue Label N

ABOVE TABLE NOT CRITICAL




1553.1000 179.70250 101549.3636 173.66190 111551.1429 172.08930 211613.5500 172.99204 101610.0000 167.16205 111611.6905 165.65533 21

Groupvisualization routineNO visualization routineTotalvisualization routineNO visualization routineTotal

Pre Test Meters Rowed

Post Test Meters Rowed



Multivariate Testsb

.891 155.794a 1.000 19.000 .000 .891

.109 155.794a 1.000 19.000 .000 .8918.200 155.794a 1.000 19.000 .000 .8918.200 155.794a 1.000 19.000 .000 .891.000 .000a 1.000 19.000 .985 .000

1.000 .000a 1.000 19.000 .985 .000.000 .000a 1.000 19.000 .985 .000.000 .000a 1.000 19.000 .985 .000

Pillai's TraceWilks' LambdaHotelling's TraceRoy's Largest RootPillai's TraceWilks' LambdaHotelling's TraceRoy's Largest Root

Effecttest

test * group

Value F Hypothesis df Error df Sig.Partial EtaSquared

Exact statistica.

Design: Intercept+group Within Subjects Design: test

b.

ABOVE TABLE IGNOREMauchly's Test of Sphericityb

Measure: distance

1.000 .000 0 . 1.000 1.000 1.000Within Subjects Effecttest

Mauchly's WApprox.


Epsilona



a.

Design: Intercept+group Within Subjects Design: test

b.





Measure: distance

38400.234 1 38400.234 155.794 .000 .89138400.234 1.000 38400.234 155.794 .000 .89138400.234 1.000 38400.234 155.794 .000 .89138400.234 1.000 38400.234 155.794 .000 .891

.091 1 .091 .000 .985 .000

.091 1.000 .091 .000 .985 .000

.091 1.000 .091 .000 .985 .000

.091 1.000 .091 .000 .985 .0004683.135 19 246.4814683.135 19.000 246.4814683.135 19.000 246.4814683.135 19.000 246.481

Sphericity AssumedGreenhouse-GeisserHuynh-FeldtLower-boundSphericity AssumedGreenhouse-GeisserHuynh-FeldtLower-boundSphericity AssumedGreenhouse-GeisserHuynh-FeldtLower-bound

Sourcetest

test * group

Error(test)


Partial EtaSquared

Look for significance in the "Within subjects" table FOR THE DEPENDENT VARIABLES OF INTEREST (INTERACTION & MAIN WITIHIN-SUBJECT EFFECT), using the appropriate line (G-G or Sphericity assumed, as explained above).

ABOVE TABLE LOOK AT (IN THIS ORDER)1. THE INTERACTION OF THE MAIN FACTORS (TEST & GROUP)IF the interaction is not sig, then proceed with the next analysis of main effect2. The main effect of the within first factor (TEST)

In this example: interaction is not sig, main effect of TEST is sig.

Tests of Within-Subjects Contrasts

Measure: distance

38400.234 1 38400.234 155.794 .000 .891.091 1 .091 .000 .985 .000

4683.135 19 246.481

testLinearLinearLinear

Sourcetesttest * groupError(test)


Partial EtaSquared

ABOVE TABLE IGNORE


Measure: distanceTransformed Variable: Average

104810222 1 104810222.3 1752.516 .000 .989139.048 1 139.048 .002 .962 .000

1136306.035 19 59805.581

SourceInterceptgroupError


Partial EtaSquared

IF the interaction above is not sig, then proceed with the next analysis of main effect



3. The main effect of the between second factor (group)

In this example: main effect of GROUP is NOT sig.

Profile Plots

1 2

test

1540.00

1560.00

1580.00

1600.00

1620.00

Estim

ated

Mar

gina

l Mea

ns

Groupvisualization routineNO visualization routine

Estimated Marginal Means of distance


Example #3: two-way ANOVA, between-within. (Jen, athlete self confidence training data)

Ask your questionJen (Graduate student 2003) asked if there was a difference in the sport related self confidence of high school runners, if the athlete was trained in using a success visualization routine. All subjects were tested for sport related self confidence (pre test). Subjects were then randomly assigned to the group that received the success visualization routine training, or did not receive the training. After 15 weeks, during which the visualization group (group 1) did visualization training plus regular team training, while the no visualization group (group 0) did regular team training, all subjects were tested again (post test). This is the same example as practiced in SPSS lesson unit 8 example #3.



Collect your data1 = treatment0= control

Subject # Group Pretest score for confidencePosttest score for confidence

1 1 38 462 1 41 413 1 39 454 1 40 445 1 39 456 1 38 477 1 38 468 1 40 509 0 41 50

10 0 36 3911 0 44 5012 0 40 4313 0 44 4714 0 39 4115 0 34 3916 0 42 3917 0 39 4018 0 39 4519 0 30 3320 0 42 4121 0 41 42

Run statistical analysis using SPSSUse same methods as in example #2 above.

Interpret SPSS output to find answer to your question. (Output tables that are not critical to our analysis are not included below)




39.3077 3.98716 1339.1250 1.12599 839.2381 3.16077 2142.2308 4.79850 1345.5000 2.56348 843.4762 4.33150 21

GROUPCONTROLTREATMENTTotalCONTROLTREATMENTTotal

PRE TEST-SPORTSELF CONFIDENCE

POST TEST-SPORTSELF CONFIDENCE



Mauchly's Test of Sphericityb

Measure: confiden

1.000 .000 0 . 1.000 1.000 1.000Within Subjects Effecttest

Mauchly's WApprox.


Epsilona



a.

Design: Intercept+GROUP Within Subjects Design: test

b.



Measure: confiden

214.077 1 214.077 42.636 .000 .692214.077 1.000 214.077 42.636 .000 .692214.077 1.000 214.077 42.636 .000 .692214.077 1.000 214.077 42.636 .000 .69229.506 1 29.506 5.876 .025 .23629.506 1.000 29.506 5.876 .025 .23629.506 1.000 29.506 5.876 .025 .23629.506 1.000 29.506 5.876 .025 .23695.399 19 5.02195.399 19.000 5.02195.399 19.000 5.02195.399 19.000 5.021

Sphericity AssumedGreenhouse-GeisserHuynh-FeldtLower-boundSphericity AssumedGreenhouse-GeisserHuynh-FeldtLower-boundSphericity AssumedGreenhouse-GeisserHuynh-FeldtLower-bound

Sourcetest

test * GROUP

Error(test)


Partial EtaSquared

Look for significance in the "Within subjects" table FOR THE DEPENDENT VARIABLES OF INTEREST (INTERACTION &



MAIN WITIHIN-SUBJECT EFFECT), using the appropriate line (G-G or Sphericity assumed, as explained above).

ABOVE TABLE LOOK AT (IN THIS ORDER)1. THE INTERACTION OF THE MAIN FACTORS (TEST & GROUP)IF the interaction is not sig, then proceed with the next analysis of main effect

In this example: interaction is sig. Because the interaction is significant the analysis stops there and you do not examine the main effects to determine if they are sig or not (because if the main effect is sig, it will be only for one of the groups and not the other, so the main effect does not reflect all the other factor groups combined, which is what the main effect is examining)


Measure: confidenTransformed Variable: Average

68368.352 1 68368.352 3045.340 .000 .99423.590 1 23.590 1.051 .318 .052

426.553 19 22.450

SourceInterceptGROUPError


Partial EtaSquared

Because the interaction above is sig, we do not use this table to examine the main effect of the groups.



Profile Plots

1 2

test

39.00

40.00

41.00

42.00

43.00

44.00

45.00

46.00

Estim

ated

Mar

gina

l Mea

ns

GROUPCONTROLTREATMENT

Estimated Marginal Means of confiden


How to do post hoc tests for the Two-way between-within measures ANOVA, following a significant interaction?

Opinions vary widely on how to do post-hoc testing in a repeated measures design, consult with you thesis chair for advice on the method to use.

Compare group means across the repeated factor using paired t tests between each of the measures, using Bonferonni correction (i.e. level of significance = 0.05/# of comparisons) (references: Using SPSS for Windows, S Green, N Salkind, T Akey, 1997, Prentice Hall, pgs 273). Compare group means across the between factor using independent t tests between each of the measures, using the Bonferonni correction. (See pg 205 Statistics in Kinesiology 2nd ed where this approach is advocated using the example of a between-within 3x3 ANOVA, so simple ANOVAs rather than t-tests are discussed, although no mention of correction of the sig p value for the repeated ANOVAs was made).

NOTES & SOURCES



post hoc mixed ANOVA 2x2in word doc: http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=6&cad=rja&ved=0CFsQFjAF&url=http%3A%2F%2Fwww.utexas.edu%2Fcourses%2Fschwab%2Fsw388r7_spring_2007%2FSolvingProblemsInSPSS%2FSolving%2520Mixed%2520Models%2520ANOVA%2520Problems.doc&ei=51-BUveiB8KhigKRtIDoDA&usg=AFQjCNElO-_euVb8ZRZ6WoWciTk3xLVwlA&sig2=jRYlRDYl2Q2OonJu_CsajA&bvm=bv.56146854,d.cGE

Pg 9: set up main effects post hocPg: 25 interpreting outputWill this work for only 2 levels of repeated measure?If so, it says for each of the 2 main effects, if the comparisons along that main effect are sig.It does not say if the 2 post-test are sig different from each other. Or 2 pre-tests sig different.________http://stats.stackexchange.com/questions/58404/post-hoc-for-2x2-mixed-design-anova-using-spssdiscusses the 2x2 mixed post hoc I ,says you never need to do a post hoc, the interaction in a 2x2 tells you all you need to know.says SPSS won't allow post hoc with less than 3 levels.____http://psychology.illinoisstate.edu/jccutti/138web/spss/spss11.htmldiscusses the 2x5 mixed post hoc,but post hoc discussion is on different example with 3 levels of repeated.

POST-HOC TESTING FOR JEN’S DATA

Use Bonferonni adjusted alpha = alpha/# comparisons= 0.05 / 4 = 0.0125

t-Test: Paired Two Sample for Means

TREATMENT Pretest score for confidence

TREATMENT Posttest score for confidence

Mean 39.125 45.5

Variance 1.267857143 6.571428571Observations 8 8Pearson Correlation -0.42068384Hypothesized Mean Difference 0df 7t Stat -5.627108301P(T<=t) one-tail 0.000396584t Critical one-tail 1.894578605P(T<=t) two-tail 0.000793168t Critical two-tail 2.364624252


http://psychology.illinoisstate.edu/jccutti/138web/spss/spss11.html

http://stats.stackexchange.com/questions/58404/post-hoc-for-2x2-mixed-design-anova-using-spss

http://stats.stackexchange.com/questions/58404/post-hoc-for-2x2-mixed-design-anova-using-spss

http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=6&cad=rja&ved=0CFsQFjAF&url=http%3A%2F%2Fwww.utexas.edu%2Fcourses%2Fschwab%2Fsw388r7_spring_2007%2FSolvingProblemsInSPSS%2FSolving%2520Mixed%2520Models%2520ANOVA%2520Problems.doc&ei=51-BUveiB8KhigKRtIDoDA&usg=AFQjCNElO-_euVb8ZRZ6WoWciTk3xLVwlA&sig2=jRYlRDYl2Q2OonJu_CsajA&bvm=bv.56146854,d.cGE






t-Test: Paired Two Sample for Means

CONTROL Pretest score

for confidenceCONTROL Posttest score for

confidenceMean 39.30769231 42.23076923Variance 15.8974359 23.02564103Observations 13 13Pearson Correlation 0.758211818Hypothesized Mean Difference 0df 12t Stat -3.347876854P(T<=t) one-tail 0.002901447t Critical one-tail 1.782287556P(T<=t) two-tail 0.005802895t Critical two-tail 2.17881283

t-Test: Two-Sample Assuming Equal Variances

TREATMENT Pretest score for confidence

CONTROL Pretest score for confidence

Mean 39.125 39.30769231Variance 1.267857143 15.8974359Observations 8 13Pooled Variance 10.50759109Hypothesized Mean Difference 0df 19t Stat -0.125422572P(T<=t) one-tail 0.450753309t Critical one-tail 1.729132812P(T<=t) two-tail 0.901506618t Critical two-tail 2.093024054

t-Test: Two-Sample Assuming Equal Variances

TREATMENT Posttest score for confidence

CONTROL Posttest score for confidence

Mean 45.5 42.23076923Variance 6.571428571 23.02564103Observations 8 13Pooled Variance 16.96356275Hypothesized Mean Difference 0df 19t Stat 1.766420374P(T<=t) one-tail 0.046693802t Critical one-tail 1.729132812P(T<=t) two-tail 0.093387603t Critical two-tail 2.093024054



Interpretation of post-hoc tests: The treatment group had a significant increase in confidence. The control group had a significant increase in confidence. There was no significant difference in confidence at the start of the

experiment, when comparing the control and treatment groups. There was no significant difference in confidence at the end of the

experiment, when comparing the control and treatment groups. (But interaction said that both groups did not respond similarly over the experiment?)

1 2

test

39.00

40.00

41.00

42.00

43.00

44.00

45.00

46.00

Estim

ated

Mar

gina

l Mea

ns

GROUPCONTROLTREATMENT

Estimated Marginal Means of confiden


Unit 9: MANOVA

ANOVA examples illustrated in previous unit are for ONE Dependent Variable in experiment – UNIVARIATE (one variable) experiment


INDEPENDENT VAR 2


Control (n=10) HRrest HRrest


HRrest HRrest

What if you have >1 DV = MULTIVARIATE experiment


INDEPENDENT VAR 2


Control (n=10) HRrestBPsystolic

HRrestBPsystolic


HRrestBPsystolic

HRrestBPsystolic


Unit 9: MANOVA

Analyse multivariate experiment using MANOVA (MultivariateANOVA)?

What is a linear combination of your DVsSport Psychol example:DV1 = self confidenceDV2 = extroversionNew DV3 = DV1 + DV2 = self confidence + extroversion (“extroverted self confidence”)

Exercise Science example:DV1 = HRrestDV2 = BPsystolicNew DV3 = DV1 + DV2 = HRrest + BPsystolic (“????????????????????”)


Does your hypothesis state that you are interested in, and will examine, linear combinations of your DVs?

YESNO

Use MANOVA Use 1 ANOVA for each DV, and apply Bonferonni correction

Unit 9: MANOVA

References for judicial use of MANOVA1) Vincent, 1999, pg 217

2) Thomas and Nelson, 2001, pg 163-4

Relationships among the dependent variable is NOT of interest to us = no use of MANOVA


Unit 9: MANOVA

Discussion of how you deal with more than one DV, using a statistical test for each DV:Source: Clinical Gait Analysis Web Page http://www.univie.ac.at/cga/faq/statistics.html, Retrieved 6-2-04We are a little green when it comes to the statistical interpretation of gait lab data. I have been attempting to look at the effects of two orthotic interventions on pathological gait. We have completed ANOVA analysis on 18 gait parameters so far. We were quite pleased with ourselves, until it was pointed out that our error level of 5%, effectively meant we had a 1 in 20 chance of interpreting a parameter as significantly different; when error was to true cause of the difference. It was suggested that it is possible to test the sensitivity of each parameter, although he was unable to shed any light on how this is accomplished. Could you provide any further enlightenment??

At least you are thinking about this issue. Its exceedingly common to see papers published who have investigated a statistical tests on a thousand different parameters and find that 20 of them show "statistically significant results" at the 5% level. It's particularly prevalent in gait analysis where there is no shortage of parameters to look at. There are various solutions. There is a related though not identical phenomenon that the more parameters you look at the higher the percentage of statistically significant results will be the product of random chance.

The most commonly used is called the Bonferroni correction which effectively says the more tests you do the lower the level at which you should accept statistical significance. Any decent stats book will guide you through this process as its applied to multiple t-tests. The principle is the same for ANOVA but I'm not sure whether the technical details are the same.

A much stronger method is to limit the number of parameters you look at before you start. Preferably nominate one key parameter in advance and stick to this - what ever you do make sure you can count the number of parameters on one hand (and no polydactyly). How you do this is up to you. You can either use you clinical skill and judgement to nominate these or do a pilot study, run tests on all the variables, and use the data to nominate the top five variables for a definitive trial. The problem with this is that in the present environment no-one will believe you. Running multiple statistical comparisons on data is so common that it will be assumed that you've done all those tests and just reported the good results. In big scale projects you can now actually pre-declare your primary outcome measures with the Lancet before you start to ensure that you don't cheat. This is a little over the top for a most of us mere mortals though.

Another approach which I've heard proposed recently from a visiting lecturer from the UK (Dr Jonathan Sterne, University of Bristol, UK - I gather he's just brought a new book out which it may be worth looking for) is to move away from assuming that anything below 5% is significant and anything above is not. Clearly there's little difference between p=0.0499 and p=0.0501 and its daft to have a precise cut-off. Sterne would have you look at the p-values as indicating comparative levels of confidence in results. This then forms the basis for a balanced assessment of the data and suggestion of probable explanations (which may include the suggestion that any particular result is a chance finding). In biomechanics it is rare that your parameters are ever fully independent and finding patterns within your significance values amongst related parameter can be powerful evidence of a real effect rather than an aberration. Using 5% as a clear cut-off makes the process of science appear objective but this is a lie. We should accept that the interpretation of results is subjective and get down to the nitty-gritty of doing this honestly and intelligently.




Unit 9: MANOVA

Another hang-up of Stern's, which is partially related, and reasonably well supported in the literature is to focus more on confidence limits in interpretting data rather than p-values.

I find Martin Bland's An introduction to medical statistics to be an excellent guide to these issues (although it is a little superficial in its treatment of ANOVA). Bland (mostly with Doug Altman) has also written a number of articles on related issues for the BMJ which can be accessed easily through his web-site (http://www.mbland.sghms.ac.uk/jmb.htm).

This whole area is a can of worms but you've got no option but to get to grip with it if you want to valid science.

Hope this is useful. Richard Baker

Gait Analysis Service Manager, Royal Children's Hospital , Flemington Road, Parkville, Victoria 3052 , Adjunct Associate Professor, Physiotherapy, La Trobe University , Honorary Senior Fellow, Mecahnical and Manufacturing Engineering, Melbourne University

Use of the Bonferonni correction:J Martin Bland, Douglas G Altman, Multiple significance tests: the Bonferroni method BMJ 1995;310:170 (21 January)http://bmj.bmjjournals.com/cgi/content/full/310/6973/170

“Why does this happen? If we test a null hypothesis which is in fact true, using 0.05 as the critical significance level, we have a probability of 0.95 of coming to a not significant--that is, correct--conclusion. If we test two independent true null hypotheses, the probability that neither test will be significant is 0.95x0.95=0.90. If we test 20 such hypotheses the probability that none will be significant is 0.9520=0.36. This gives a probability of 1-0.36=0.64 of getting at least one significant result--we are more likely to get one than not. The expected number of spurious significant results is 20x0.05=1. In general, if we have (kappa) independent significant tests at the (alpha) level of null hypotheses which are all true, the probability that we will get no significant differences is (1-(alpha))(kappa). If we make (alpha) small enough we can make the probability

that none of the separate tests is significant equal to 0.95. Then if any of the (kappa) tests has a P value less than (alpha) we will have a significant difference between the treatments at the 0.05 level. Since (alpha) will be very small, it can be shown that (1-(alpha))(kappa)(about)1-(kappa) (alpha). If

we put (kappa)(alpha)=0.05, so (alpha)=0.05/(kappa), we will have probability 0.05 that one of the (kappa) tests will have a P value less than (alpha) if the null hypotheses are true. Thus, if in a clinical trial we compare two treatments within five subsets of patients the treatments will be significantly different at the 0.05 level if there is a P value less than 0.01 within any of the subsets. This is the Bonferroni method. Note that they are not significant at the 0.01 level, but at only the 0.05 level.”


http://bmj.bmjjournals.com/cgi/content/full/310/6973/170

mailto:[email protected]

http://www.mbland.sghms.ac.uk/jmb.htm

Unit 9: MANOVA

Above final concept also stated by: Knudson, D., (2009) Significant and meaningful effects in sports biomechanics research. Sports Biomechanics, 8(1) ,96 – 104. (see pg 99)


http://www.informaworld.com/smpp/title~content=g908693897~db=all

http://www.informaworld.com/smpp/title~content=t776628940~db=all~tab=issueslist~branches=8#v8

http://www.informaworld.com/smpp/title~content=t776628940~db=all

Unit 9: MANOVA

Demonstration of WHY you need to do a correction to your critical value of α if you are doing more than one statistical test in your research project.

If you have multiple tests, and there are REALY no sig differences for any of the tests…# of tests

Probability (% chance) of concluding 1 or more false positives by chance =(100(1-0.95n)

1 52 103 144 195 23

10 4015 5420 64

Better way to correct alpha for multiple comparisons….Bonferonni correction is conservative

“Conservative” in statistics means that you have a tendency to NOT reject the null and conclude you do NOT have evidence of a sig difference between groups, when in fact you should.

Holm’s correction to control for type I & II errors while maintaining statistical power, when doing multiple statistical tests.6

The hypotheses tested are arranged from the smallest to the largest observed P-value. For comparisons (i) from 1 to n, each observed P-value is compared with the adjusted critical P-value according to the formula:Pc = α / (n – i + 1)

Pc:critical P-valueα: 0.05n: total number of comparisonsi: number of THIS comparison in the sequence from low to high

6 Knudson, D., (2009) Significant and meaningful effects in sports biomechanics research. Sports Biomechanics, 8(1) ,96 – 104.


http://www.informaworld.com/smpp/title~content=g908693897~db=all

http://www.informaworld.com/smpp/title~content=t776628940~db=all~tab=issueslist~branches=8#v8



Unit 10: Measurements

1) RELIABILITY= consistent measures= score will be identical if two measurements are made

2) OBJECTIVITY (eliminate subjectivity)= two different individuals giving the test would arrive at the same score for the same subject

OBJECTIVITY DEPENDS ON:• clarity of scoring system

HOW DO YOU GET OBJECTIVITY, WITH A SCORING SYSTEM WHICH IS NOT TOTALLY CLEAR?

3) VALIDITY= the test measures what it is suppose to measure

VALIDITY DEPENDS ON:• the test being reliable• the test being relevant

4 TYPES OF RELEVANCE LOGICAL CONCURRENT

predictive construct

LOGICAL VALIDITY

subjective (expert) decision that the test measures what it claims toeg., 50-yd dash is a test of running speed

CONCURRENT VALIDITYyou have a valid method to measure somethingsee if the measures made by a new test correlate well to the established standard


Unit 11: Experimental Design

From: Field, KB et al., (2007) Should Athletes Stretch Before Exercise? Sports Science Exchange (SSE # 104), 20(1). Retrieved 7-5-07 from http://www.gssiweb.com/Article_Detail.aspx?articleid=736&level=4&topic=4

Stretching and Injury Prevention

“…there must be many high-quality studies of stretching and injury prevention in the literature”

Herbert and Gabriel (2002) found only two studies that would qualify under a standard criterion for methodological quality.Thacker et al. (2004) performed a comprehensive search of literature on stretching for injury prevention, including articles published through 2002. Using a different quality assessment tool than that employed by Herbert and Gabriel, they screened 361 articles and found only six they deemed of sufficient quality to merit analysis.

Why are there so few studies that are of sufficient quality to be considered to contribute to the topic?

You want to be able to argue that any differences in dependent variable between groups is due to your treatment (independent variable)

INTERNAL VALIDITYthe extent to which the results can be attributed to the TREATMENTS used in the study

EXTERNAL VALIDITYthe generalizability of the results of the study

Paradox: the more you control (to increase the internal validity) the LESS generalizable the conclusion become (less externally valid)

EIGHT THREATS TO INTERNAL VALIDITY

1) HISTORYEvents occurring during the experiment which are not part of the treatmente.g., do some subjects participate in exercises, in addition to the ones in your research program?

2) MATURATIONProcesses within the subjects, that operate as a result of time passing (e.g., aging, hunger,

fatigue)e.g., do people score better on a post-test because they are older (bigger, stronger)?

3) TESTINGthe effects of one test on subsequent administration of the same testse.g., do people score better on a post-test because they learned during the first test?

4) INSTRUMENTATIONchanges in instrument calibration, causing lack of agreement within or between observerse.g., does an instrument change with time & use, including the use of people as observers

(observer drift)?


http://www.gssiweb.com/Article_Detail.aspx?articleid=736&level=4&topic=4


5) STATISTICAL REGRESSIONthe fact that groups selected on the basis of extreme scores are not as extreme on subsequent

testinge.g., if you group people based on a score (e.g., high, medium, low), in a second test the extreme

groups will have moved toward the middle. Due to selection of extremes of scores, not averages which people tend to over time.

6) SELECTION BIASidentification of comparison groups in other than a random mannere.g., if groups are not randomly formed, then any differences observed after treatment could be

due to differences that existed between the groups before the treatment, not the treatment.

Showing that the groups are not different before the test on the dependent variable (or any other measure) does NOT overcome nonrandom grouping, there may be something other than what you measure different between the groups.

7) EXPERIMENTAL MORTALITYloss of subjects for nonrandom reasonse.g., do people in one group quit more than another group, due to the treatment they receive?

e.g., The control group is bored & quits, or the treatment group does not like testing & quits.

8) EXPECTANCYtesters anticipating (unconsciously usually) that certain subjects will do better than otherse.g., are the “skilled” & “unskilled” subjects rated the same? Are post tests expected to better in

the treatment, but not control, groups

FOUR THREATS TO EXTERNAL VALIDITY

1) REACTIVE OR INTERACTIVE EFFECTS OF TESTINGThe pretest may make the subject more sensitive to the treatment, so the treatment is not as

effective without the preteste.g., the person realizes from pretest he is in very poor shape, and works hard during treatment.

2) INTERACTION OF SELECTION BIASES AND THE EXPERIMENTAL TREATMENTA group is selected on a characteristic, so treatment works only on groups with that characteristice.g., you selected to work on elite athletes, the psychological techniques work well on elite

athletes, but not on lower levels

3) REACTIVE EFFECTS OF EXPERIMENTAL ARRANGEMENTStreatments are effective in the testing situation, but not in other situationse.g., do subjects behave the same way in the testing situation, as they do in other settings?Hawthorne effect: subjects’ performance changes when attention is paid to them

4) MULTIPLE TREATMENT INTERFERENCEif more than one treatment is given, the earlier treatment(s) influence subsequent onese.g., is a drug or fatigue effect of the first test eliminated before second test?

CONTROLLING THREATS TO INTERNAL VALIDITY

RANDOM ASSIGNMENT TO GROUPS:equating the subjects in the experimental & control groups, allows the assumption that the groups

do not differ at the start of the experimentcontrols for:

HISTORY up to the start of exp.



MATURATIONSTATISTICAL REGRESSIONSELECTION BIASES

For within subjects designs:subject serves as own control & receives both control & experimental treatment(s)RANDOM ASSIGNMENT OF TREATMENT ORDER (some receive control condition first, some

receive treatment(s) first)controls for:

TESTING

e.g., control group (C) & one treatment (T)C-TT-C

e.g., control group (C) & two treatments (T1, T2 )6 possible combinations to randomly assign subjects to

C - T1 - T2

C - T2 - T1

T1 - C - T2

T1 - T2- CT2- C - T1

T2 - T1 - C

PLACEBOS, BLIND & DOUBLE BLIND SETUPS:control for:

HAWTHORNEEXPECTANCYAVIS EFFECT (subjects in control group try harder because they are in the control group)

Placebo: used to evaluate if treatment effect is real or psychological. Subject receives same attention & interaction with experimenter

Bind setup: Subject does not know which treatment group subjects are in(not always possible)

Double-Bind setup: Subject & Experimenter do not know which treatment group subjects are in

THREATS TO INTERNAL VALIDITY NOT CONTROLLED BY ABOVE METHODS:

Instrumentation: ensure you have valid & reliable testing methods

Experimental Mortality : ways to prevent subject loss?• clear explanation of project• use select population that is already committed to participate in

team activities, exp. becomes part of activities (reduces external validity)

CONTROLLING THREATS TO EXTERNAL VALIDITY

NB: Ensuring the subjects, treatments, tests, and testing situations represent larger populations / situations

random sampling, (or “good enough” sampling of subjects, treatments) is critical

TYPES OF EXPERIMENTAL DESIGNS



Nomenclature:R = random assignment of subjectsO = observation or test, (O1, O2, ...)T = treatment applied (T1, T2, ...)

blank space = control (no treatment)....... = groups not randomly assigned

1) ONE SHOT STUDYT O1Uselessperformance at observation can not be attributed to treatment

2) ONE GROUP, PRETEST - POSTTEST DESIGNO1 T O2Essentially useless (Pilot study ?)

If people are better at O2:is it due to: treatment OR (due to lack of control group)…maturation, history (before or during the experiment), testing effect, selection bias (if group was

not randomly selected from population)

3) STATIC GROUP COMPARISONT O1..................

O2

Compare two groups, one receives the treatment, one does notBUT: subjects decide which group they will be in (not random)

e.g. people volunteer to go in exercise group or control group.Essentially useless (Pilot study ?)If O1 O2: is it due to: treatment, history (before or during the experiment)



FOLLOWING DESIGNS (4-6) ARE TRUE EXPERIMENTAL DESIGNS due to randomization

4) RANDOMIZED GROUPSR T1 O1R O2Analyze: using independent t-test, O1 versus O2to answer the question: “ Does a treatment produced change?”

e.g., training or no training groups, O = aerobic capacitysubjects randomly assigned to treatment or control groups

If you control for threats to internal validity not controlled by randomization (threats controlled by randomization listed above), then differences between O1 & O2 are due to treatment (T)

Multiple treatments possibleR T1 O1R T2 O2



R O3

e.g., 2x/week, 3X/week training or no training groupsAnalyze: using ANOVA, O1 versus O2 versus O3

5) PRETEST-POSTTEST RANDOMIZED GROUPSR O1 T1 O2R O3 O4

to answer the question: “What is the AMOUNT of change produced by a treatment”Analyze: using factorial repeated measures ANOVA

6) SOLOMON FOUR-GROUPR O1 T1 O2R O3 O4R T1 O5R O6

Only design to specifically evaluate REACTIVE OR INTERACTIVE EFFECTS OF TESTING threat to external validity

because you can compare the equivalent groups, with & without the pretestdifficult to analyzedifficult to recruit enough subjects for

FOLLOWING DESIGNS ARE QUASI-EXPERIMENTAL DESIGNS

7) EX POST FACTOSimilar to design 3, but experimenter did not apply the treatmente.g., test athletes & nonathletes to look for variables which distinguish the two groups.

E.g. #2: Note: This study includes it’s OWN control group (not comparing to “control” data elsewhere in the literature)

Sig. results can lead to ideas to test using experimental design, but design can not prove that any different characteristics found influenced the two groups to be different



e.g. Follow-up study

E.g. #3: Do not just describe a sample from a population. Note: This study includes it’s OWN control group (not comparing to “control” data elsewhere in the literature)

8) EPIDEMIOLOGYCase studies - studies of individuals with a disorderCohort studies - study of large groups, some of whom have been exposed to the cause of

disorder in questionretrospective, prospective or cross-sectional designs

9) SINGLE SUBJECT DESIGNTIME SERIESO1 O2 O3 O4 T1 O5 O6 O7 O8

baseline, or constant rate of change is established O1-O4, O5-O8rate of change is different O4-O5

line A, B, C, treatment effect indicated, D, E, F, G no treatment effect



analysis : repeated measures ANOVA does not always work (e.g., line C above)regression techniques to test slopes and intercepts better

REVERSAL DESIGNbaseline, then treatment, then no treatment, then treatment, then no treatmentO1 O2 T1 O3 O4 T2 O5 O6

Also designated: ABABAA = baseline condition B = treatment condition

line A, B, C, treatment effect indicated, D, E, F, G no treatment effect

analysis :regression techniques to test slopes and intercepts

some possible questions addressed by single subject designs: does the treatment produce the same effect each time? are the effects of the treatment cumulative, or does measurement return to baseline after

treatment? do varying intensities, frequencies, & lengths of treatments produce varying responses?

SUMMARY (unit 506/11)Threats to Internal Validity

HIS

TOR

Y

MA

TUR

ATI

ON

TES

TIN

G

INS

TRU

ME

NTA

TIO

N

STA

TIS

TIC

AL

RE

GR

ES

SIO

N

SE

LEC

TIO

N B

IAS

EX

PE

RIM

EN

TAL

MO

RTA

LITY

EX

PE

CTA

NC

Y

R to treatment and control

R, and occurs in both treatment and control groups

R, Pre & post test in bothtreatment and control groups

DB, & control for intrasession history



XXXX drop outs, even if equal in control and treatment groups

P,B,DB



Threats to external ValidityR

EA

CTI

VE

OR

IN

TER

AC

TIV

E

EFF

EC

TS O

F TE

STI

NG

INTE

RA

CTI

ON

OF

SE

LEC

TIO

N B

IAS

ES

A

ND

TH

E

EX

PE

RIM

EN

TAL

TRE

ATM

EN

T

RE

AC

TIV

E

EFF

EC

TS O

F E

XP

ER

IME

NTA

L A

RR

AN

GE

ME

NTS

MU

LTIP

LE

TRE

ATM

EN

T IN

TER

FER

EN

CE

Because the interaction between taking a pretest and the treatment itself may affect the results of the experimental group, it is desirable to use a design which does not use a pretest

Although selection is controlled for by randomly assigning subjects into experimental and control groups, there remains a possibility that the effects demonstrated hold true only for that population from which the experimental and control groups were selected.

Often can not avoid artificiality of the experimental setting and the subject's knowledge that he is participating in an experiment

R = Random assignmentPBDB = Placebos, Blind & Double Blind setups


SPSS Lesson

Download the Excel data file for this lesson from the KIN 506 web site: PE506SPSS_lesson_data.xls

Before you can use SPSS you must use Excel to set up data the way that SPSS requires it be organized. The way in which you should set up your data in SPSS may not be the same as in Excel. Each row is one case (or subject). Column is a variable. Use grouping variables to distinguish cases which are different groups. e.g., gender (category variable), treatments (treatment variables). Repeated measures on the same subject are different variables (i.e., put in different columns).

START SPSSSelect the option to “type in data” to open a new empty data sheet

Select and copy all the data for a test to be run in excel, do not include the header cells with the column labels.

Place the cursor into the top left cell in the SPSS data sheet and paste the data in.

Click on the “Variable View” tab at the bottom and name the variables. Eight characters or less, no spaces or special characters, always start with a character (not a number), Set # decimals, ADD “LABEL” TO EACH VARIABLE TO MAKE OUTPUT EASY TO READ, for a grouping variable set the “values” for the variable (i.e. define the groups),

Do:

Unit 7 SPSS example of one-way, repeated measures ANOVA

Unit 8 SPSS example of two-way, between-between ANOVA

Unit 8 SPSS example #1 of two-way, between-within ANOVA

Unit 8 SPSS example #2 of two-way, between-within ANOVA


Systematic Literature Search Assignment

KIN 506 Systematic Literature Search (SLS) Assignment

Why do we need to search so many sources??? Testimony from a former student.Woo hoo! Thanks for changing the link to Science Direct....because today, on the last search for the SLS assignment, was a journal article there I have never seen before.....with my exact transversus abs test on Pilates participants. I have never seen anyone take that test and apply it to Pilates or yoga, and finding a very recent article which did that is about the greatest thing I could imagine in terms of making my thesis "go". I had never heard of the journal it was in, and the great thing is that the journal appears to be a perfect fit for an article with my thesis results in it. The study I found today is cross sectional, while mine is a training study, and mine will be stronger I think. Anyway, I am telling you all of this because I know there are reasons you give us these assignments, and I am bouncing of the walls to have discovered something so important and critical, especially when it didn't come up in so many other searches. You give these great lectures on why in-depth searching is important, and now I am a believer (not that I didn't believe you before, but I wasn't quite as sold as I am now!)Print the two files in the SLS assignment files list on the course web pageSave a copy of the fifth file in the list (KIN506SLS.doc)How to complete the library search, and how to cite in APA style will be discussed in class. Below is a copy of the file KIN506SLS.doc that will be discussed. Note that this copy is for discussion in class, work the assignment directly into the KIN506SLS.doc file you download.

Name: Topic:

Systematic Literature Search Assignment for KIN 506You are to look up your topic in each of the databases listed. When you find a relevant source in a database, give the complete and accurate bibliographic citation, using APA style. Use the APA Manual (6th Edition). If you cannot find anything on your topic in a particular reference after going back at least five years, then look for older information. Only circle "Not Found" and leave the citation space blank if you cannot find any information published at any time.You cannot cite the same article twice, even if it appears in different databasesDatabases to search. The links to access these databases are found in the KIN 506 web page related to this assignment – http://myweb.facstaff.wwu.edu/chalmers/506.html

1. Science DirectAPA Citation Not Found


Systematic Literature Search Assignment

2. EBSCO (WWU Database list: E, “EBSCO – Multiple Database Search“) Once you are in the database, you must “Choose Databases” (link at center top), de-select “Business Source Complete, and select “ERIC”, Health Source: Nursing/Academic Edition”, “Military & Government Collection”, “SPORTDiscus”, “CINAHL Complete”, PLUS if you are a sport psychology student include “PsycARTICLES”, “PsycINFO”

APA Citation Not Found

3. Catalog of US Government Publications (Government Printing Office) http://catalog.gpo.gov/ (Use advanced search mode)


4. PubMedAPA Citation Not Found

5. One search result from one of the three following databasesI. PEDro - The Physiotherapy Evidence Database

II. REHABDATA Disability and rehabilitation researchIII. CIRRIE Disability and rehabilitation research


6. Google Scholar (For the KIN 506 assignment, try to find a source that is not from Medline)


7. WWU OneSearchhttp://onesearch.library.wwu.edu/primo_library/libweb/action/search.do?mode=Advanced&vid=WWU&tab=everythingThe above link is to the advanced search mode, ENSURE YOU ARE IN THE TAB AT THE TOP FOR "+Summit + Articles", THEN, limit format (in box a right) to "Articles", THEN do search.


8. Stanford University HighwireAPA Citation Not Found

9. ProQuest American and international dissertations published by University Microfilms International



http://onesearch.library.wwu.edu/primo_library/libweb/action/search.do?mode=Advanced&vid=WWU&tab=everything

http://onesearch.library.wwu.edu/primo_library/libweb/action/search.do?mode=Advanced&vid=WWU&tab=everything

http://catalog.gpo.gov/

Term Paper Assignment

KIN 506 Term paper assignmentWritten Report

Complete a research proposal on a topic of interest following the PEHR Guidelines for a thesis that are being discussed in class (see the proposal format in Master’s Degree Candidate Booklet, and first page of lecture handouts). Your literature search assignment will (hopefully) form the basis of the research needed for the term paper. Use at least 15 research articles in your review (Chapter II), although you may use many times this minimum. The use of non peer reviewed sources is rare in scholarly research in our field, so use it judiciously if at all, and justified if you use them. This includes the use of internet sources that are not peer reviewed. The only internet sources that are acceptable are ones that provide scholarly information that is not available elsewhere. You can not use undergraduate or graduate text books as references. You must ensure that you provide a reference for all statements you make in your paper which are not your own original thought. For references you provide for this assignment, citation of the class lecture notes as a reference is not allowed.

The paper must be written using APA format, with the methods in past tense (for easy conversion to a thesis). APA style extends beyond the reference list style. All aspects of the paper formatting are defined. See chapter 5 (Manuscript preparation and sample paper) of the APA guide, 6th edition, and the details to support chapter 5 in the other parts of the guide. The term paper grading sheet and the thesis checklist (the latter filled in by you) are to be included with the term paper, please staple these onto the report after the last page of the references. A sample of the thesis checklist follows the blank one

Ensure that your paper is checked by a human before submission. A paper that is only checked by a spelling checking program may contain grammatical errors, or errors of disjointed thought or poor organization. Only human checking by a person other than the original author can detect these sorts of errors.

Written report due date: The due date is listed in the course outline. Papers submitted after start of class on this date lose 10% of the maximum total marks per day, including each weekend day.

Extensions and incomplete grades: Extensions and incomplete grades will be allowed only for medical reasons or very significant personal reasons.

Oral ReportA presentation of your research and practice proposal will be scheduled

for the last week of the term. Prepare a verbal summary of your proposal for the class. Your presentation must include visual aids (power point or similar presentation software) to assist in your presentation of material. The duration of the report will be determined during the term, based on the number of people who need to present, but will be approximately ten



minutes. Your presentation must include references to key studies discussed, placed after a stated fact.



KIN 506 Proposal Checklist

Title:

List of all independent/treatment variables:

Null Hypothesis (Circle the names of each of the independent variables(s), and underline each of the dependent variables(s)):

Study design (give name):

Is this a true experimental design? (Y/N)List EACH dependent variable (DV)

The way you obtain this DV is listed in methods?

The statistical analysis of this DV is in statistics description?

Y / N Y / NY / N Y / NY / N Y / NY / N Y / NY / N Y / NY / N Y / NY / N Y / NY / N Y / NY / N Y / NY / N Y / N



SAMPLE: KIN 506 Proposal Checklist

Title: H-reflex size in healthy elderly and young adults

List of all independent/treatment variables:

Age, step cycle phase

Null Hypothesis (Circle each of the independent variables(s), and underline each of the dependent variables(s)):

There is no difference in the size

of the H-reflex in the different

phases of the step cycle when

comparing young and elderly

adults

Study design (give name): Ex post facto

Is this a true experimental design? (Y/N)

No

List EACH dependent variable (DV)

The way you obtain this DV is listed in methods?

The statistical analysis of this DV is in statistics description?

H-reflex size during walking Y / N Y / N

standing H-reflex size Y / N Y / N

walking speed Y / N Y / N

Y / N Y / N

Y / N Y / N

Y / N Y / N


Term Paper Grading Sheet

KIN 506 Term Paper Grading

NAME: TOPIC

SCALE:Exceptional

Obvious extra &

outstanding effort in

research & reporting

GoodCorrect facts at

sufficient detail to explain issues, and

written presentation without any

significant errors

AdequateFacts & detail explain

issues, but more detail, easily possible, or some explanation confusing. Written

presentation with some significant errors

WeakFacts & detail do

not explain issues, or some significant

errors. Written presentation with very significant

errors

Nee

ds H

elp!

5 4 3 2 1Chapter I: The Research

Problem5 4 3 2 1 Comments

(A) Introduction

(B) Purpose

(C) Hypothesis

(D) Significance of the study

(E) Limitations of Study

(F) Definition of TermsChapter II: Literature Review

A) Introduction & Summary

B) Body of Review

Chapter III: Methods

(A) Brief Introduction

(B) Description of study population

(C) Design of the study


Term Paper Grading Sheet

(D) Data collection procedures

D1. Instruments

D2. Discussion of techniques & procedures

D3. Data processing &/or subject training (optional)

(E) Data (statistical) analysisPREPARATION OF PAPER

Paper and paragraph organization of material, use of headings to organize material

Spelling, punctuation, & sentence structure

APA style reference list. At least 15 research articles, no undergraduate or graduate text books, use of non-peer reviewed sources must be justified.

Correct use of citations in text

General Comments:


Term Paper Presentation Grading

NAME: TOPIC

GoodCorrect facts at sufficient detail to

explain issues, written and oral presentation without any significant

errors

WeakFacts & detail do not explain

issues, or some significant errors. Written and/or oral presentation

with very significant errors

MISSING!

2 1 0Chapter I: The Research

Problem2 1 0 Comments

Purpose

Hypothesis - Identify IV & DV

Significance of the study

Chapter II: Literature Review

Review –approx 50-60% of total time (including references)

Chapter III: Methods

Methods description

PREPARATION OF TALK

Effective use of visuals

Voice clear, Looking at audience


Term Paper Presentation Grading

Did not exceed time limit of 10 ±2 min, or need to be cut off, but not excessively short under time limit either:________

General Comments:


Introduction to statistics assignments & assignment. #1

INTRODUCTION TO THE STATISTICS ASSIGNMENTS

To do this, and all the statistics assignments, you will need to downloaded from the KIN 506 web site the excel data file 506_Asst_data.xlsx. The file contains variables measured on undergraduate classes, and copied from previous thesis research in the PEHR dept. One column is one data set variable. Each row, within the same gender and within one of the subject columns, is all the data for one individual. The name for each variable is abbreviated in the excel file, the full description of each variable is shown in the following table. Scroll to the right to find all the data.

For assignments 1-3:VARIABLE NAME DESCRIPTION OF VARIABLE

MB 490 490 male biceps curl max data (lbs)FB 490 490 female biceps curl max data (lbs)

MW 490 490 male weight data (lbs)FW 490 490 female weight data (lbs)

MBG 490 490 male biceps girth data (inches)FBG 490 490 female biceps girth data (inches)OLDFBF Old female biceps figures/data (lbs lifted, 1 RM, dominant arm)right, left right and left hand grip strength, pairs of measurements in a number

of different people (lbs)T1VO2, T2VO2, T3VO2 test 1,2,3 of repeated measures on VO2 (ml/kg/min) on subjects

listedtest1 = control, test2 = pizza diet, test3 = ice cream diet

For assignments 4-6, the variables are described and defined in the assignment.

FOR ALL ASSIGNMENTS: You are to work in pairs (or groups of 3 if the class is large). The people working together will submit one report with both names on it. The goal is to learn how to use the computers, and how to do statistical analysis. You will learn more (and faster) if you put your heads together and teach each other, than if you do it on your own. Two groups submitting the same, or essentially the same, material is plagiarism.

REPORT REQUIREMENTS: You are to complete the following tasks using the data provided. Each task is to be done using Excel unless the assignment specifies you to use SPSS. You are to submit a print out of the Excel or SPSS result table for those tasks labeled “Show Excel/SPSS result table”, or report the value calculated by excel/SPSS for tasks labeled “Report the value”. When presenting a result table from SPSS, highlight the specific critical lines or values in the SPSS output that you read when interpreting the output. Also include the required discussion of the statistical analyses. You do not print out your whole table of student data for the submission of the assignment. Reports are due at the start of class on the due date announced when the assignment is distributed. All elements of your report are to be included in a single Word document that is printed. The report must follow the sequence of tasked listed, and each part must be labeled (i.e., put the label "Task 2a" just before you display you results for that task). The plots you produce must meet the following criterion:

Ensure plots do not include background shading, or grid lines. Plots with a single set of data must not have a legend, plots containing two sets of

data must have a legend.


Introduction to statistics assignments & assignment. #1

Scale on each axis of a plot must be set so that the plot does not have excessive wasted space along the axis where there is no data.

The scale marker labels should have decimals only if this degree of precision is present in the data being graphed.

The steps in the units labeling an axis should be fine enough that values can be read from the graph, but not so fine that the graph axis looks cluttered. For example, on a graph axis that has a minimum of 0 and a maximum of 100, having three markers at 0, 50, 100 is using too large of steps in the axis labels. Using axis labels at 0, 5, 10, 15,…,90, 95, 100, will likely be too cluttered. Find a happy medium that is not cluttered, but gives resolution for the reader.

Axis labels, in the cases of the data used here, should not have decimals. Each axis must have a line in black (not grey) and tick marks at the labels. Text in the graph must be black (not grey). Error bars on bar charts must be easily visible from the patterns that fills the

bars.

STATISTICS ASSIGNMENT #1 Descriptive StatisticsUse EXCEL for this assignment.

TASK VARIABLES (data columns) TO USE

TASK #1, For each of the data sets: (use functions or formulas only)

Compute the mean, standard deviation, minimum & maximum (Report the values)7

FB 490OLDFBF

TASK #2, For each of the data sets:Plot a bar graph that displays the mean and standard deviation error bars for the males and females, for both body weight and weight lifted. The x-axis is to have the two groupings of "Body Weight" and "Biceps Weight Lifted", the legend is to distinguish the male and female data.

MB490FB490MW490FW90

You must do all calculations within your excel spreadsheet, using built in functions, or formulae you write.Reports are due at the start of class on the due date announced in class. You submit BOTH a printed copy of your requirements, and you email to Dr. Chalmers the excel file used to produce the report.

7 In your introductory statistics course you should have learned also to test each variable to ensure it is normally distributed. For your thesis data, you can do this with SPSS using: analyze => descriptive statistics => explore: Select your dependent variables and move them to the Dependent list box. Then select the “Plots” button, and select the “Normality plots with tests” option. In the output look for the “Tests of Normality” box, and use the Shapiro-Wilk test to determine if the data is normally distributed. It is not normal if sig < 0.05. . (Source: Discovering Statistics using SPSS for Windows, (2000) Andy Field, Sage, London., pp. 46 -47, 50 - 51).


Statistics assignment #2 - Correlation & regression

You are to submit a print out of the specified EXCEL output for each task, and the required discussion points. For all tests of significance use a level of = 0.05. Use Excel for this assignment.

TASK VARIABLES TO USE

For each of the data set pairs:a) Draw a scattergram showing the relationship between the two

variables, with girth on the x-axis and strength on the y-axis. Do one graph that contains both of the pairs of data. Place a regression line on the graph for each of the two pairs of data. Ensure that the two lines use obviously different line patterns so they are easily distinguished when printed.

b) Calculate the correlation between the two variables, for both of the pairs of data (i.e., you will have two correlation values). (Report the values)

c) For both the males and the females, calculate if the correlation is significant. Report: the degrees of freedom used, the critical value of = 0.05, and your conclusion regarding the significance of the correlation.

pair #1

MB 490

MBG 490

pair #2

FB 490

FBG 490



Statistics assignment #3 Comparing two groups

Complete the following tasks using EXCEL8. For all tests of significance use a level of = 0.05.


TASK #1, for the pair of variables listed:

a) Use a T-test to test the means of the biceps maximum of the male and female groups. (Show the entire Excel result table including title line, but not the student data. Use the analysis tools not the function t-test)

b) Calculate the meaningfulness of the difference in the means using both the effect size (using pooled standard deviation) and the omega square formulae.

c) How do you interpret the results in parts a - b? (i.e., Say in words what the statistical results are telling you as you would in the results section of your thesis).

FB 490

MB 490

TASK #2, for the pair of variables listed:

a) Use a T-test to test the means of the grip strength of the two hands. (Show entire Excel result table including title line, but not the student data. Use the analysis tools not the function t-test)

b) Calculate the meaningfulness of the difference in the means using both the effect size (using pooled standard deviation) and the omega squared formulae.

c) How do you interpret the results in parts a - b? (i.e., Say in words what the statistical results are telling you as you would in the results section of your thesis).

right

left

(subject set 1 in the assignment data file)


8 For this assignment we will use Excel for a t-test because our emphasis here is on calculating and interpreting the meaningfulness of the differences observed. For a thesis, it is better to use SPSS, as the SPSS t-test procedure does a test for the assumption of equal variances in the two groups, and runs the t-test for both possibilities.


Statistics assignment #4 One-way, within group ANOVA

Complete the following tasks using SPSS. For all tests of significance use a level of = 0.05.


TASK #1, for the variables listed:

a) Use an ANOVA to test the means of the three tests on the runners (Subject set 2 in the assignment data file). The same runners each ran on 3 different types of treadmills (1) Control (2) a very stiff deck (3) a very flexible deck (the very flexible deck is thought to make the running less jarring, but may take more energy), and VO2 max was measured under each condition. Show only the SPSS result tables used in the interpretation of the output, and highlight the key values used in your interpretations. Also include the plot of estimate marginal means.

b) To determine if the results are meaningful, report the Eta Squared value in the Within-Subjects effects table, and interpret using the Vincent 1999 guidelines.


c) How do you interpret all of the results in parts a - b? (i.e., Say in words what the statistical results are telling you as you would in the results section of your thesis).

T1VO2 (Control deck)

T2VO2 (a very stiff deck)

T3VO2 (very flexible deck)

Reports are due at the start of class on the due date announced in class. You submit a printed copy of this assignment.


Statistics assignment #5 Two-way ANOVA, Between-Between


Shenelle asked if there was a difference in level of exercise (Kcals expended per week in moderate or greater intensity exercise) performed by people following their cardiac event (e.g., heart attack) if they participated in the St. Joe’s cardiac rehabilitation program, versus if they did not participate in the program. She surveyed people after their heart attacks at two different time points, at times: 1 (= first measure at 5-7 months post cardiac event) and 3 (= third measure at 23-25 months post cardiac event). To get the data at the two time points she did not survey the same people, she surveyed (in the same month) different people who were in the appropriate time frames, post heart attack.


TASK #1, for the variables listed:

a) Use a two-way ANOVA to test the effects of group (participant or non participant in cardiac rehabilitation) and time post cardiac event, on level of exercise (Kcals expended per week in moderate or greater intensity exercise) performed. Show only the SPSS result tables used in the interpretation of the output, and highlight the key values used in your interpretations. Also include the plot of estimate marginal means.

b) To determine if the results are meaningful, report the Eta Squared value for the interaction, and each of the main effects, and interpret using the Vincent 1999 guidelines.


c) Report the p value (called “sig” in SPSS output) and whether the hypothesis tested is statistically significant or not, for the interaction and each of the main effect tests.

d) How do you interpret all of the results in parts a - c? (i.e., Say in words what the statistical results are telling you as you would in the results section of your thesis).

Data set labeled for assignment #5 in the assignment data file



Statistics assignment #6 Two-way ANOVA, Between-Within


TASK 1: April asked if there was a difference in maximum heart rate while rowing on an ergometer in a 5 minute ergometer rowing competition, if an athlete was trained in using a precompetition visualization routine. All subjects were tested for maximum heart rate during 5 minutes of an ergometer rowing competition (pre test). Subjects were then randomly assigned to the group that received the visualization training, or did not receive the training. After 12 weeks, during which the visualization group (group 1) did visualization training plus regular rowing team training, while the no visualization group (group 2) did regular rowing team training, all subjects were tested again (post test).

TASK 1 VARIABLES TO USE

a) Use a two-way ANOVA to test the effects of group and time (pretest time and posttest) on maximum heart rate while rowing on an ergometer in a 5 minute ergometer rowing competition. Show only the SPSS result tables used in the interpretation of the output (including the table of descriptive statistics), and highlight the key values used in your interpretations. Also include the plot of estimate marginal means.

b) To determine if the results are meaningful, report the Eta Squared value for the interaction, and each of the main effects, and interpret using the Vincent 1999 guidelines.


c) Report the p value (called “sig” in SPSS output) and whether the hypothesis (or hypotheses) tested is (are) statistically significant or not, for the interaction and each of the main effect tests.

d) Perform post-hoc tests of the repeated measures main effect, if needed. Use excel and show the complete t-Test tables that you used. Include the interpretation of the post-hoc tests in part e below.

e) How do you interpret all of the results in parts a - c? (i.e., Say in words what the statistical results are telling you as you would in the results section of your thesis).

Data set labeled for assignment #6a in the assignment data file

Assignment continues on the next page….


Statistics assignment #6 Two-way ANOVA, Between-Within

TASK #2: Melissa (Graduate student 2004) asked if there was a difference in the throwing velocity of college pitchers, if the athlete was trained with over weight balls. All subjects were tested for throwing velocity (pre test). Subjects were then randomly assigned to the group that received the training with over weight balls, or did not receive the training with over weight balls. After 18 weeks, during which the over weight balls group (group 1) did over weight ball training plus regular pitcher training, while the no over weight ball group (group 1) did regular pitcher training, all subjects were tested again (post test).

TASK 2 VARIABLES TO USE

a) Use a two-way ANOVA to test the effects of group and time (pretest time and posttest) on throwing velocity. Show only the SPSS result tables and graphs used in the interpretation of the output (including the table of descriptive statistics), and highlight the key values used in your interpretations. Also include the plot of estimate marginal means.

b) To determine if the results are meaningful, report the Eta Squared value for the interaction, and each of the main effects. NOTE: FOR THIS DATA SET ONLY REPORT THE ETA SQUARED VALUE AND IT’S INTERPRETATION FOR EACH HYPOTHESIS TESTED IF IT SHOULD BE REPORTED, BASED ON THE INTERPRETATION OF THE DATA. Interpret using the Vincent 1999 guidelines.


c) Report the p value (called “sig” in SPSS output) and whether the hypothesis (or hypotheses) tested is (are) statistically significant or not, for the interaction and each of the main effect tests. NOTE: FOR THIS DATA SET ONLY REPORT THE P VALUE AND IT’S INTERPRETATION FOR EACH HYPOTHESIS TESTED IF IT SHOULD BE REPORTED, BASED ON THE INTERPRETATION OF THE DATA.

d) Perform post-hoc tests of the repeated measures main effect, if needed. Use excel and show the complete t-Test tables that you used. Include the interpretation of the post-hoc tests in part e below.

e) How do you interpret all of the results in parts a - c? (i.e., Say in words what the statistical results are telling you as you would in the results section of your thesis).

Data set labeled for assignment #6b in the assignment data file



Documents

myweb.facstaff.wwu.edumyweb.facstaff.wwu.edu/~chalmers/ElichMaterials/Elich_Course_… · Web viewKIN 506. Research Design in Human Movement and Performance. Lecture notes, Fall