Latent Variable Models for Misclassified Polytomous Outcome Variables

Latent variable modeling is a multivariate statistical technique commonly used in the social and behavioral sciences to describe an underlying structure which cannot be measured directly. The models used in such analysis relate all observed variables to latent common factors. The analysis is traditionally carried out under the assumption that the observed variables are continuous with a multivariate normal distribution. In many studies, the response variables are in polytomous form which are often affected by misclassification errors. In this paper, we propose a new latent variable modeling approach which takes into account the response error associated with the measured polytomous outcome variables. An EM algorithm is proposed for parameter estimation. To validate the benefits of our approach, a simulation study is conducted. The approach is illustrated using a substance abuse prevention study.

By: Jens C. Eickhoff, Yasuo Amemiya

Published in: Lecture Notes in Computer Science, volume 4276, (no ), pages 1772-89 in 2006

Please obtain a copy of this paper from your local library. IBM cannot distribute this paper externally.

Questions about this service can be mailed to .