Analysis of data arising from a stratified design with the cluster as unit of randomization

Area of Science:

• Biostatistics
• Clinical Trials
• Epidemiology

Background:

• Cluster randomization is used in various research designs.
• Analyzing dichotomous outcomes in stratified cluster designs presents challenges.
• Assessing overall intervention significance across strata is crucial.

Purpose of the Study:

• To develop statistical methods for analyzing dichotomous data from stratified cluster randomization.
• To assess the statistical significance of intervention effects in such designs.

Main Methods:

• Proposes a weighted paired t-test.
• Utilizes the empirical logistic transform.
• Applies to designs with random assignment of two clusters per stratum.

Main Results:

• The proposed method provides a way to assess intervention significance.
• The technique is suitable for designs involving large aggregate clusters within strata.

Conclusions:

• The weighted paired t-test offers a robust approach for dichotomous data analysis in stratified cluster randomization.
• This method enhances the evaluation of intervention effects in complex study designs.