Five interval estimators for proportion ratio under a stratified randomized clinical trial with noncompliance

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Statistical Inference

Background:

• Randomized clinical trials (RCTs) often involve patient noncompliance and confounding variables.
• Accurate estimation of treatment effects, specifically the proportion ratio (PR), is crucial in stratified RCTs.
• Existing methods may not adequately address challenges posed by noncompliance within stratified designs.

Purpose of the Study:

• To develop and evaluate interval estimators for the proportion ratio (PR) in stratified RCTs with noncompliance.
• To compare the finite sample performance of five proposed asymptotic interval estimators.
• To provide practical guidelines for selecting appropriate estimators based on study characteristics.

Main Methods:

• Development of five asymptotic interval estimators for the PR.
• Inclusion of estimators based on weighted-least squares (WLS), Mantel-Haenszel weights, and Fieller's Theorem.
• Application of Monte Carlo simulations to assess coverage probability and average length.
• Consideration of optimal weights for different estimation approaches.

Main Results:

• The study evaluates the coverage probability and average length of five distinct interval estimators.
• Performance comparison across various scenarios reveals differences in estimator efficiency and accuracy.
• Limitations and usefulness of each estimator are discussed in detail.

Conclusions:

• The research provides a comparative analysis of interval estimators for PR in complex RCT settings.
• Findings guide the selection of appropriate statistical methods for analyzing stratified RCTs with noncompliance.
• The study contributes to robust statistical inference in clinical trial research.