The eirt program is an add-in for Excel that does Item Response Theory estimations. This program is restricted to the case where the latent variable (the ability) is unidimensional and follow the normal distribution, and the number of possible response (the options) is finite. This type of item include the binary case, the multiple choice case and the graded case.
The list of supported models includes the logistics models with one, two and three parameters for the binary data, the nominal response model for the multiple choice data, and the graded response model for the graded data.
The estimation of the item parameters is done by the maximum marginal likelihood estimator or by the Baye modal estimator. In addition, two non-parametrics estimators are supported; the kernel estimator and the penalized maximum marginal likelihood estimator. The abilities are estimated by the expected a posteriori estimator.
The eirt program uses the numerical routines from the libirt library (http://libirt.sf.net) also available in rirt a R package (http://www.r-project.org).
