Applying Open Source to Adaptive Testing

Embed Size (px)

Citation preview

  • 7/24/2019 Applying Open Source to Adaptive Testing

    1/25

    Applying Open Source to

    Adaptive TestingMichal [email protected]

    TAO DAYS Berlin, 10/09/2012

    www.psychometrics.cam.ac.uk

    Michal Kosinski

    [email protected]

  • 7/24/2019 Applying Open Source to Adaptive Testing

    2/25

    PART I: Open-source

    PART II: Item Response Theory and Computerized

    Adaptive Testing

    PART III: What do you need to build your CAT (or anyother test or e-learning platform)

    and where to get a right open-source software

    PLAN

  • 7/24/2019 Applying Open Source to Adaptive Testing

    3/25

    Part I:

    Open-source

    Michal Kosinski

    [email protected]

  • 7/24/2019 Applying Open Source to Adaptive Testing

    4/25

    Who is not convinced to open-source software?

    Problems? Concerns?

    Using open-source

  • 7/24/2019 Applying Open Source to Adaptive Testing

    5/25

    High reliability / free updates / constant debugging

    Open source gives you independence: Own your software copy and access all the updates

    Enhance your software as (and for how much) you wish

    See what is under the hood

    Low switching costs = innovation and progress

    Usually relatively low costs

    Why open-source instead ofcommercial one?

  • 7/24/2019 Applying Open Source to Adaptive Testing

    6/25

    all the aforementioned advantages, plus:

    Avoids rediscovering the wheel

    Minimize dependence on your suppliers / employees

    Decrease maintenance costs (e.g. debugging /updating)

    Open-source elements as a part ofyour own software dev. project

  • 7/24/2019 Applying Open Source to Adaptive Testing

    7/25

    It doesnt mean that you have to stop offering commercials0lutions (e.g. mySQL model)

    Low-cost marketing Build trust

    Get free input from the community

    Ideas / Improvements / Enhancements / Debugging

    Increased opportunities for funding

    Improve the world! (and get competitive edge!)

    Publish your stuff as open-source

  • 7/24/2019 Applying Open Source to Adaptive Testing

    8/25

    How do I sell products based on open-source?

    In the same way as commercial ones!

    Most open-source can be used for commercial purposes

    Open-source based products based can be sold / licensed / etc.

    Who do I blame/sue when something goes wrong?

    Read your agreement with a commercial software developercan you really blame them?

    Choose your software wisely Pay someone to give you a warranty

    Problems and myths

  • 7/24/2019 Applying Open Source to Adaptive Testing

    9/25

    Part II:Item Response Theory

    and ComputerizedAdaptive Testing

    www.psychometrics.cam.ac.uk

    Michal Kosinski

    [email protected]

  • 7/24/2019 Applying Open Source to Adaptive Testing

    10/25

    Modelling on the item level (and not on the test level) Robust item parameters

    Can be used for test development and scoring

    Score is item independent

    Advantages:

    Increased accuracy of the tests

    Cheaper test development

    On-going test development Test security / IP protection

    Enables Adaptive Testing (CAT)

    IRT

  • 7/24/2019 Applying Open Source to Adaptive Testing

    11/25

    Select items fit for a person and purpose

    Advantages:

    Much shorter tests

    Less frustration / more attention

    Detect react to inattention, cheating, etc.

    Tests fit for purpose (e.g. screening)

    Further increased test security and protection

    CAT

  • 7/24/2019 Applying Open Source to Adaptive Testing

    12/25

    Adaptive learning

    Innovative question formats

    Incorporate non-question information

    IRT and CAT beyond testing!

  • 7/24/2019 Applying Open Source to Adaptive Testing

    13/25

    Part III:

    What do you need to buildyour CAT (or any other testor learning process)

    www.psychometrics.cam.ac.uk

    Michal Kosinski

    [email protected]

  • 7/24/2019 Applying Open Source to Adaptive Testing

    14/25

    1. Build a test = develop an Item bank Item writing / translations

    2. Establish the psychometric properties of the test

    CAT / IRT analysis

    3. Choose the most appropriate CAT approach

    CAT simulation

    4. Deliver your test on the testing platform

    5. Analyse the results

    What you need:

  • 7/24/2019 Applying Open Source to Adaptive Testing

    15/25

    Consider open source items: Avoid rediscovering the wheel!

    IPIP.orgi.org

    myPersonality.org/wiki

    Crowd-source your items / translations e.g. Amazon Mechanical Turk

    Wiki-source your items?

    Generate your items automatically:

    Prof. Heinz Hollings

    Cambridge IQ items generator (coming soon!)

    Build an item bank

  • 7/24/2019 Applying Open Source to Adaptive Testing

    16/25

    R language for statistical analysis

    Classical and IRT analyses (+ anything else you may

    want) ltm, irttoys, psych, and loads of other psychometric

    libraries

    Powerful graphing capabilities

    Flexible programming interface

    Great wealth of analyses available within oneenvironment

    Establish the psychometricproperties of the test

  • 7/24/2019 Applying Open Source to Adaptive Testing

    17/25

    Issues:

    Where to start?

    How to select next item?

    When to stop?

    What is the expected accuracy and test length?

    CAT simulation :

    catR package

    catIrt

    MAT (Multidimensional Adaptive Testing, M3PL model)

    Choose the most appropriate CATapproach

  • 7/24/2019 Applying Open Source to Adaptive Testing

    18/25

    Use web-based technology:

    Compatibility

    Can be used locally

    Plan ahead:

    Choose the platform on the piloting stage

    Use CAT/IRT enabled platform even if do you need it yet

    Testing platform choice

  • 7/24/2019 Applying Open Source to Adaptive Testing

    19/25

    It is open source

    It relies on open-source software (e.g. R, MySQL)

    Ease of use develop advanced tests without a softwareengineer

    Flexible:

    Robust R engine

    Works with any CAT / IRT algorithm

    Nicely integrates with other software e.g. Moodle, TAO, etc.

    Concerto:an open source adaptive testing

    platform

  • 7/24/2019 Applying Open Source to Adaptive Testing

    20/25

    Design of Concerto

    Concerto Platform

    Database

    MySQL

    Statistical

    Engine

    R server

    Item Editor

    CKEditorQTI

    Editor

    Open-source elements Flexible R engine

    Standard language inpsychometrics

    Reliable / up-to-date methods Powerful MySQL database Modular design

    Add / Replace modules ifnecessary

  • 7/24/2019 Applying Open Source to Adaptive Testing

    21/25

    Internet

    Secured LAN e.g. testing centre

    Locally e.g. Tablet with local copy of the platform

    Administration channels

  • 7/24/2019 Applying Open Source to Adaptive Testing

    22/25

    Concerto: brief tour

  • 7/24/2019 Applying Open Source to Adaptive Testing

    23/25

    Ror

    GNU Octave

    Analyse your results:

  • 7/24/2019 Applying Open Source to Adaptive Testing

    24/25

    Avoid rediscovering the wheel Do your homework BEFORE you decide to buy/develop

    software

    Stay independent: Insist on using open-source software EVEN if you decide

    to develop your own platform

    Use IRT and CAT it is cheaper, more accurate, andgives you a competitive edge!

    Insist on using R in statistical analyses

    Help us to fully integrate TAO with Concerto Anyone has spare programmer or few euros?

    TAKE-HOME MESSAGES

  • 7/24/2019 Applying Open Source to Adaptive Testing

    25/25

    More about Concerto:www.psychometrics.cam.ac.uk

    Michal Kosinski, [email protected]

    Prof. John Rust, [email protected]

    www.psychometrics.cam.ac.uk

    Michal Kosinski

    [email protected]