Assessing Classification Bias in Latent Class Analysis: Comparing Resubstitution and Leave-One-Out Methods

Authors

Marc H. Kroopnick, Association of American Medical CollegesFollow
Jinsong Chen, The George Washington UniversityFollow
Jaehwa Choi, The George Washington UniversityFollow
C. Mitchell Dayton, University of MarylandFollow

Abstract

This Monte Carlo simulation study assessed the degree of classification success associated with resubstitution methods in latent class analysis (LCA) and compared those results to those of the leaveone- out (L-O-O) method for computing classification success. Specifically, this study considered a latent class model with two classes, dichotomous manifest variables, restricted conditional probabilities for each latent class and relatively small sample sizes. The performance of resubstitution and L-O-O methods on the lambda classification index was assessed by examining the degree of bias.

DOI

10.22237/jmasm/1272686760