Rasch models for measuring change

This article presents Rasch models for the analysis of longitudinal discrete data and testing specific hypotheses about change. It is shown that Rasch models can be represented in terms of log-linear models and hierarchical linear or factor models with a logit link function, which allows the specification of Rasch models using general software packages. The use of Rasch models for longitudinal data is illustrated with an empirical example. The final section discusses model extensions for the analysis of polytomous items, multidimensional latent constructs, and items with varying discrimination parameters. Keywords: Rasch model; latent trait model; change; longitudinal research; log-linear model; hierarchical linear model; factor model; link function