In Search of Multivariate Associations: Comparison of CHAID, Log-Linear Analysis and Multiple Correspondence Analysis

This study is intended to justify the need and demonstrate the possibilities of searching for multivariate associations of categorical variables, which are the most common in quantitative sociological research. Despite the presence of both theoretical and empirical arguments in favor of the multivariate analysis of such variables, the practice of using appropriate methods is still weakly widespread – partly because there is no universal method for such analysis, as there are no recent works that generalize approaches to it. In this study, we have tried to fill this gap by comparing three methods: CHAID, log-linear analysis, and multiple correspondence analysis. The comparison of methods took place at the theoretical and empirical levels. According to the results of the study, the descriptive and prognostic capabilities of the methods differ significantly: multiple correspondence analysis obtained combinations that are the most saturated regarding interpretation, and log-linear analysis obtained the most useful for forecasting combinations. The latter conclusion contradicts the current opinion about the efficiency of the CHAID in cases of a target variable presented in the data and therefore has a high practical significance for the further development of the idea of building high-precision predictive models in sociological research.