Some general methods for the analysis of categorical data in longitudinal studies

Area of Science:

• Biostatistics
• Epidemiology
• Clinical Research

Background:

• Longitudinal studies generate complex multivariate categorical data.
• Analyzing such data, especially with sparse frequencies, presents statistical challenges.
• Existing methods may not adequately address hypotheses in repeated measures designs.

Purpose of the Study:

• To present methods for analyzing multivariate categorical data in longitudinal epidemiologic and clinical studies.
• To develop and illustrate large-sample tests for pertinent hypotheses using weighted least squares and Wald statistics.
• To explore strategies for handling sparse frequency data in repeated measurement designs, including a generalized Mantel-Haenszel approach.

Main Methods:

• Application of weighted least squares to generate Wald statistics for hypothesis testing.
• Development of a generalized Mantel-Haenszel strategy for tests of marginal homogeneity (symmetry) in repeated measures.
• Utilizing SAS software (version 5), specifically the FREQ and CATMOD procedures.

Main Results:

• Demonstration of Wald statistics for hypothesis testing with multivariate categorical data.
• Illustration of the generalized Mantel-Haenszel strategy for ordinal repeated measures data.
• Successful implementation of statistical methods within SAS computing software.

Conclusions:

• The proposed methods provide effective tools for analyzing longitudinal multivariate categorical data.
• The generalized Mantel-Haenszel strategy offers a viable approach for sparse data in repeated measures.
• SAS software facilitates the application of these advanced statistical techniques in practice.