mirt-package: Full information maximum likelihood estimation of IRT models.
Description
Full information maximum likelihood estimation of
multidimensional IRT modelsDetails
Analysis of dichotomous and polytomous response data
using latent trait models under the Item Response Theory
paradigm. Includes univariate and multivariate one-,
two-, three-, and four-parameter logistic models, graded
response models, generalized partial credit models,
nominal models, multiple choice models, and multivariate
partially-compensatory models. These can be used in an
exploratory or confirmatory manner with optional user
defined linear constraints. Exploratory models can be
estimated via quadrature or stochastic methods, a
generalized confirmatory bi-factor analysis is included,
and confirmatory models can be fit with a
Metropolis-Hastings Robbins-Monro algorithm which can
include polynomial or product constructed latent traits.