The use of GEE for analyzing longitudinal binomial data: a primer using data from a tobacco intervention

Area of Science:

• Behavioral Science
• Addiction Research
• Biostatistics

Background:

• Longitudinal study designs are crucial in addiction research for tracking changes over time.
• Intervention studies often involve repeated measurements, leading to within-participant correlation.
• Accurate analysis of longitudinal data requires accounting for this correlation, which can be complex.

Purpose of the Study:

• To provide an overview of Generalized Estimating Equations (GEE) for analyzing correlated longitudinal data.
• To highlight the utility of GEE for addiction researchers, particularly for non-normal data.
• To demonstrate GEE application using a tobacco smoking relapse prevention intervention study.

Main Methods:

• Overview of Generalized Estimating Equations (GEE) methodology.
• Application of GEE to analyze correlated binary data from a longitudinal intervention study.
• Focus on accounting for within-participant correlation in repeated measurements.

Main Results:

• GEE is a valuable tool for analyzing correlated longitudinal data in addiction research.
• The method effectively handles non-normal data, common in behavioral studies.
• Demonstrated successful application in a smoking relapse prevention context.

Conclusions:

• Generalized Estimating Equations (GEE) offer a robust approach for analyzing complex longitudinal data in addiction research.
• Encourages wider adoption of GEE by behavioral researchers to improve analytical rigor.
• Highlights the importance of appropriate statistical methods for intervention effectiveness studies.