---

MODELS | METHODS | LICENSE | DOWNLOAD | APPLICATIONS | DOCUMENTATION | AUTHORS | THANKS | REFERENCES | FRANÇAIS

---

The Item Response Theory Library (libirt) is a set of functions written in C to estimate the items and abilities from the responses of subjects to a questionnaire.

While two command line programs ("irt" and "mirt") are also provided, the library is mostly being used in others applications.

MODELS

• 1PLM - One parameter logistic model. The threshold (difficulty or location parameter) of each item is estimated while the slope (discrimination or scale parameter) of each item can be fixed.
• 2PLM - Two parameters logistic model. Both the threshold and slope of each item are estimated.
• 3PLM - Three parameters logistic model. The threshold, slope and asymptote (guessing parameter) are estimated.
• Bock's nominal response model - The threshold and slope of each option are estimated.
• Samejima's graded response model - The modal threshold of each option and the slope of each item are estimated.
• Nonparametric - In addition to the parametric models, libirt can estimate the response functions by smoothing (see the methods section).

METHODS

The parametric estimators are:

• MMLE - Marginal Maximum Likelihood Estimator.
• BME - Bayes Modal Estimator.

The nonparametric estimators are:

• Kernel smoothing - The Nadaraya-Watson regression used in TestGraf.
• PMMLE - Penalized Marginal Maximum Likelihood Estimator.

The abilities estimators are:

• EAP - Expected A Posteriori.
• WMLE - Warm's Weighted Maximum Likelihood Estimator.

LICENSE

libirt is developed as a free software, released under the terms of the GNU Public License (GPL).

DOWNLOAD

The latest version is available here.

APPLICATIONS

Applications using this library:

• rirt - A package for R. Download available here.
• eirt - An add-in for Excel. Download available here.

DOCUMENTATION

The developper's documentation is available online. The latest version can be built from the source using Doxygen.

AUTHORS

• Stephane Germain <stephane@altsync.net>
• Pierre Valois <pierre.valois@fse.ulaval.ca>
• Belkacem Abdous <belkacem.abdous@msp.ulaval.ca>

THANKS

libirt uses the numerical routines of the GNU Scientific Library (GSL).

REFERENCES

For information concerning the models and methods:

• Baker, F.B. & Kim, S.-H. (2004). Item response theory: parameter estimation techniques. Second Edition. Dekker, New York.
• Ramsay, J.O. (1991). Kernel smoothing approaches to nonparametric item characteristic curve estimation. Biometrika, 56, 611-630.