A Latent Variable Model for Ordinal Variables

A full-information maximum likelihood method for fitting a multidimensional latent variable model to a set of ordinal observed variables is discussed. This method is an implementation of a general class of models for ordinal variables, and for regression models with one ordinal dependent variable and all explanatory variables observed. Estimation of the model, scoring of persons on the latent dimensions, and the goodness-of-fit of the model are also discussed. The method is applied to an example dataset concerning attitudes toward technology.