Notes on testing equality and interval estimation in Poisson frequency data under a three-treatment three-period...

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Methods

Background:

• Crossover designs are efficient for comparing treatments, especially in longitudinal studies.
• Analyzing event frequency data often requires specialized statistical approaches, particularly when event counts follow a Poisson distribution.

Purpose of the Study:

• To develop and evaluate statistical procedures for testing treatment equality and estimating mean frequency ratios in three-treatment, three-period crossover trials.
• To assess the performance of proposed methods using Monte Carlo simulations under various scenarios.

Main Methods:

• Development of test procedures and interval estimators for the ratio of mean frequencies assuming a Poisson distribution.
• Utilizing weighted-least-squares (WLS) and Mantel-Haenszel (MH) approaches for analysis.
• Evaluation through Monte Carlo simulations to assess Type I error rates and interval precision.

Main Results:

• Proposed test procedures maintain good Type I error control even with moderate sample sizes.
• Weighted-least-squares (WLS) test procedures demonstrate superior performance compared to common contingency table methods.
• Both WLS and Mantel-Haenszel (MH) interval estimators exhibit comparable precision in estimating mean frequency ratios.

Conclusions:

• The developed WLS-based statistical methods are effective for analyzing event frequency data in three-treatment, three-period crossover trials.
• These methods offer reliable performance for treatment comparisons and provide precise interval estimates for mean frequency ratios.
• The study provides practical tools for analyzing asthma exacerbation data, as illustrated by a clinical trial example.