by Ton Heinen
Part of the Advanced Quantitative Techniques in the Social Sciences Series
The standard latent class model, which has been popular among social scientists as an instrument for data reduction, is also a flexible tool for analyzing structural relationships between categorical variables and can be seen as a natural extension of the log-linear model in order to take measurement error into account. Among behavioural scientists latent trait models have been proposed as the preferable psychometric tools for measuring abilities in such a way that characteristics of items and individuals could be studied separately. However, what are the similarities and differences between the latent class model and latent trait models? Through a careful examination of these issues, Ton Heinen explores topics such as: how to estimate the parameters of latent class analysis models and latent trait models; methods for model selection; and ways to examine the correspondence between discrete latent trait models and certain restricted latent class models. In addition, he reviews log-linear models, latent trait models, a number of restricted latent class models, as well as procedures for the estimation of parameters for these models.