Homogeneity test of relative risk ratios for stratified bilateral data under different algorithms

Area of Science:

• Biostatistics
• Medical Statistics
• Clinical Trial Design

Background:

• Medical clinical studies frequently utilize stratified bilateral data from paired body parts.
• Ignoring the correlation between paired body parts can lead to biased or misleading research findings.

Purpose of the Study:

• To evaluate the equality of relative risk ratios across strata using optimal algorithms.
• To compare the performance of different algorithms for obtaining global and constrained maximum likelihood estimations (MLEs).

Main Methods:

• Development and application of optimal algorithms for global and constrained maximum likelihood estimations (MLEs).
• Proposal of three asymptotic test statistics: $$, $$, and $$.
• Utilizing Monte Carlo simulations to assess algorithm performance based on mean square errors and convergence rates.

Main Results:

• The Fisher scoring algorithm generally outperforms other methods, showing effective convergence for global MLEs and lower mean square error for constrained MLEs.
• Comparison of the three proposed test statistics revealed that $$ offers robust Type I error rates (TIEs) and satisfactory statistical power.
• Empirical results highlight the importance of algorithm selection in biostatistical analysis.

Conclusions:

• The Fisher scoring algorithm is recommended for its efficiency in estimating maximum likelihood.
• $$ is the preferred test statistic due to its reliable performance in controlling Type I errors and maintaining statistical power.
• Accurate statistical methodologies are crucial for reliable medical research involving bilateral data.