A1029
Title: Efficient computation for latent class analysis with missing data
Authors: Masahiro Kuroda - Okayama University of Science (Japan) [presenting]
Abstract: Latent class analysis is a statistical method for clustering and classifying multivariate categorical data. It is widely applied in the social, behavioral, and health sciences. In a latent class model, manifest variables are observed and are assumed to arise from an underlying unobserved latent variable. The latent variable represents the hidden structure or groupings within the observed data. Latent class analysis is considered, where some of the data for the manifest variable contains missing parts. The missing data mechanism is assumed to be completely missing at random (MCAR) for the latent variable and missing at random (MAR) for the manifest variables. Thus, the presence of two distinct missing data mechanisms complicates parameter estimation in the analysis. The aim is to propose an efficient estimation algorithm using a partial imputation technique for the latent class analysis with such missing data. Numerical experiments demonstrate the computational efficiency of the proposed algorithm.
