Consistent Multi-Stage Latent Class Regression with Accommodation for Differential Item Functioning

• Activity Number: 603
• Type: Topic Contributed
• Date/Time: Thursday, August 4, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Health Policy Statistics
• Author(s): Karen Bandeen-Roche*+
• Companies: The Johns Hopkins University

Abstract

Latent variable models have long been utilized by behavioral scientists to summarize constructs that are represented by multiple variables or are difficult to measure, such as health practices and psychiatric syndromes. They are regarded as particularly useful when measurable variables are imperfect surrogates that may differentially measure the construct of inferential interest; among numerous criticisms, they are criticized as being overly abstract and computationally intensive. We propose a new strategy for developing latent measurement model-based "indices" for subsequent use in regression modeling attending particularly to the potential for differential measurement (or, "item functioning") within levels of covariate values. We show that, unlike most existing such strategies, it yields consistent estimators and associated inferences for regression parameters vis a vis full latent variable regression. Small sample performance properties are evaluated. It is hoped that, by counter-balancing strengths and weaknesses of latent variable modeling, the findings will improve the utility of latent variable based approaches for scientific investigations.