Latent Class Analysis with Covariates: Applications for Public Opinion Research

• In Event: Contemporary Issues in Public Opinion Research Methods

Abstract

Latent class (LC) analysis is a model-based clustering approach for categorical data, with a wide range of applications in political science. This analytical approach is particularly useful for the study of public opinion with a focus on cross-national and sub-group heterogeneity. This paper introduces the R package multilevLCA, which is the first R package to implement stepwise estimation of multilevel LC models with covariates. When the data have a hierarchical structure, the multilevel LC model can be used to account for higher-level dependencies between the units by means of a second categorical LC variable at the group level. The research interest of single-level and multilevel LC analysis typically lies in the relationship between the LCs and external covariates, or predictors. To estimate LC models with covariates, researchers can use the statistically advantageous full-information maximum-likelihood, or one-step, approach, or the generally recommended two-step estimator, which separates the estimation of the clustering model from the subsequent estimation of the regression model. The two-step approach simplifies model interpretation and improves computation time. Yet the lack of software solutions for two-step estimation of multilevel LC models has limited the dissemination of this approaches in the applied literature. The analytical approach introduced in this paper allows for the implementation of the most comprehensive set of model specifications and estimation approaches in R, estimating single- and multilevel LC models, with and without covariates, using the one-step and two-step approaches.

Authors

• Johan Lyrvall, University of Catania

• Roberto Di Mari, University of Catania

• Zsuzsa Bakk

• Jennifer Oser, Ben-Gurion University of the Negev in Israel

• Jouni Kuha, London School of Economics