Analyzing longitudinal categorical data: Sources of uncertainty and sample methods

Summary and discussion

The purpose of this paper was to first provide an overview of methods for analysis of categorical variables in longitudinal settings and to, second, give examples of applications of such methods. The overview was presented within a framework of six problems pertinent to statistical inference (3, 13). The overview provided two perspectives. First, variable-oriented approaches were introduced. Examples were given for trend analysis using log-linear models. Second, person-oriented approaches were introduced. Specifically, methods of Configural Frequency Analysis were introduced that allow researchers to detect types and antitypes in cross-classifications.

The overview and the examples suggest that methods for analysis of longitudinal categorical data are most flexible and allow researchers to answer a wide array of questions. Three propositions, presented before the examples, posit that statistical methods for analysis of categorical data do not fall behind methods for continuous data.