Hidden Markov Models: An Approach to Sequence Analysis in Population Studies

Danilo Bolano, University of Geneva

In this paper we provide an overview of Hidden Markov Models for population studies. We will show the relevance of latent models for life course studies despite they are only limited used in literature. Starting from the general version of the model we discuss some particularly interesting extensions for social sciences providing several practical examples from national panel studies both using categorical variables like health status, life satisfactions and continuous variables like income levels. For the latter case, we present a mixture model that we have developed for analyzing longitudinal data with non homogeneous behaviors. In this model the observed heterogeneity can be induced by one or several latent factors and each level of these factors is related to a different component of the observed process. The time series is then seen as a weighted mixture and the relation between successive components is governed by a Markovian latent transition process