Latent Regression in Loglinear Rasch Models

Abstract

This article presents a framework for the use of latent variables as outcomes in regression analysis. Based on loglinear Rasch models where item parameters are known or estimated using conditional maximum likelihood a simple and fast estimation algorithm is proposed. The interpretation of regression parameters in the presence of random effects is discussed. A regression model based on a loglinear Rasch model is used to model general and specific group differences in an occupational health study.

Key Words:

• Loglinear Rasch models
• Latent regression
• Generalized linear models
• Random effects

Acknowledgments

The authors would like to thank Niels Keiding for valuable comments to a previous version of this paper. This research was supported by the Danish Research Academy.