Michael J. Zickar, Ph.D.

Department of Psychology, Bowling Green State University

mzickar@bgnet.bgsu.edu

Possibly one of the most frustrating things about IRT for first-time users is that the standard statistical packages (SPSS & SAS) cannot run the necessary analyses.  Stand-alone programs are necessary to estimate IRT models.  To complicate matters, different programs are needed to estimate different models and to handle different types of data.  There is no single, one-size-fits all IRT estimation program.  Below are some basic descriptions of various software packages.  Most of these packages are available from Assessment Systems Corporated.  Their website is located at www.assess.com.

For an excellent features comparison of these software packages, please check out this website:  http://www.assess.com/compprog.html

BILOG-MG

   Models:  Rasch, 2PL, 3PL

   Features:    Can conduct DIF analyses.

BILOG for Windows

   Models:  Rasch, 2PL, 3PL

   Features:    Windows based.  Can generate nice graphs of item response functions.

MULTILOG

   Models:  Rasch, 2PL, 3PL, Samejima's Graded Response Model, Bock's Nominal Model

   Features:    Estimates a large number of polytomous IRT models.  Can test nested models and analyze multiple groups for DIF analyses.

PARELLA

   Models:  ????   (Anybody help us here!).

   Features:  ????

PARSCALE

   Models:  Rasch, 2PL, 3PL, Samejima's Graded Response Model, Bock's Nominal Model

   Features:    Can estimate Muraki's partial credit model.

WINMIRA

   Models:  Polytomous and dichotomous Rasch models combined with latent trait models

   Features:  Windows-based, extremely easy to run.  Generates nice graphics.  Only program that combines

Annotated Bibliography

Besides the program test manuals, there are very few good sources on how to run these programs and how to handle practical problems that arise when running these programs.

Zickar, M.J. (in press).  Modeling data with polytomous item response theory.  In N. Schmitt & F. Drasgow (Eds.), New Advances in Psychometric Methods.   Jossey-Bass.

This chapter has some information on how to run MULTILOG and includes sample output with a discussion of output.  For an "in press" copy of this book chapter, please e-mail mzickar@bgnet.bgsu.edu. Steven Reise and Nils Waller have written a chapter in this book on dichotomous IRT which should include similar discussion of practical details (though I have not seen their chapter).