ABSTRACT
This chapter explores the Configural Frequency Analysis (CFA), focuses on subjects or observations. Similar to prediction analysis (PA), variable relationships are of interest only in that a log-linear base model must be defined, relative to which subjects can be described. Social science concepts covary not only for empirical but also for conceptual reasons. The terms 'clinical depression' and 'sadness' are sample cases of covarying terms. Chance models for CFA determine expected cell frequencies. Global CFA models do not form groups of variables that can be distinguished by type of relationship. Design options in CFA include group comparisons. Many statistical methods, including analysis of variance (ANOVA), require that variances be homogeneous for proper application. In the analysis of repeated observations, requirements often include homogeneity of variances over time. CFA of differences provides information concerning the direction a time series is taking by comparing time-adjacent measures.
