A Tutorial on Multiblock Discriminant Correspondence Analysis (MUDICA): A New Method for Analyzing Discourse Data From Clinical Populations

Abstract

Purpose In communication disorders research, clinical groups are frequently described based on patterns of performance, but researchers often study only a few participants described by many quantitative and qualitative variables. These data are difficult to handle with standard inferential tools (e.g., analysis of variance or factor analysis) whose assumptions are unfit for these data. This article presents multiblock discriminant correspondence analysis (MUDICA), which is a recent method that can handle datasets not suited for standard inferential techniques. Method MUDICA is illustrated with clinical data examining conversational trouble-source repair and topic maintenance in dementia of the Alzheimer’s type (DAT). Seventeen DAT participant/spouse dyads (6 controls, 5 participants with early DAT, 6 participants with moderate DAT) produced spontaneous conversations analyzed for co-occurrence of trouble-source repair and topic maintenance variables. Results MUDICA found that trouble-source repair sequences and topic transitions are associated and that patterns of performance in the DAT groups differed significantly from those in the control group. Conclusion MUDICA is ideally suited to analyze language and discourse data in communication disorders because it (a) can identify and predict clinical group membership based on patterns of performance, (b) can accommodate few participants and many variables, (c) can be used with categorical data, and (d) adds the rigor of inferential statistics.