Model Comparison and Simulation for Hierarchical Models: Analyzing Rural-Urban Migration in Thailand

Abstract

Sociological research often examines the effects of social context with hierarchical models. In these applications, individuals are nested in social contexts—like school classes, neighborhoods or villages—whose effects are thought to shape individual outcomes. Although applications of hierarchical models are common in sociology, analysis usually focuses on inference for fixed parameters. Researchers seldom study model fit or examine aggregate patterns of variation implied by model parameters. We present an analysis of Thai migration data, in which survey respondents are nested within villages and report annual migration information. We study a variety of hierarchical models, investigating model fit with DIC and posterior predictive statistics. We also describe a simulation to study how different initial distributions of migration across villages produce increasing inter-village inequality in migration.