Multiblock discriminant correspondence analysis: Exploring group differences with structured categorical data

Abstract

Psychological research often involves complex datasets that cannot easily be analyzed using traditional statistical methods. Multiblock Discriminant Correspondence Analysis (multiblock dica, also called mudica) examines group differences in large, structured categorical datasets and identifies blocks of variables that contribute to these differences. Data for this illustration were obtained from a study on mental health literacy (N = 648) that included 33 questions that were arranged into four blocks: etiology, symptoms, treatment, and general knowledge of psychological disorders. With non-parametric inference tests and results displayed as intuitive maps, mudica revealed differences in performance across groups not readily detectable using standard methods.