Common risk difference test and interval estimation of risk difference for stratified bilateral correlated data

Area of Science:

• Biostatistics
• Medical Statistics
• Correlated Data Analysis

Background:

• Bilateral correlated data are common in medical research, especially in ophthalmology and otolaryngology.
• Paired organ measurements exhibit high intra-class correlation, necessitating specialized statistical methods.
• Ignoring correlation or confounders can bias statistical inferences.

Purpose of the Study:

• To propose and evaluate statistical approaches for testing common risk differences in stratified bilateral correlated data.
• To develop confidence intervals for the common difference of two proportions.
• To address the dual challenges of intra-class correlation and confounding effects in paired data.

Main Methods:

• Three novel approaches for testing common risk differences in stratified bilateral correlated data were developed.
• The study derived five distinct confidence intervals for the common difference between two proportions.
• Monte Carlo simulations were employed to assess the performance of the proposed methods.

Main Results:

• The score test statistic exhibited robust type I error rates and high statistical power compared to other methods.
• The score confidence interval demonstrated satisfactory coverage probabilities and reasonable interval widths.
• The proposed methodologies were illustrated using a real-world otolaryngologic study dataset.

Conclusions:

• The score test statistic is recommended for analyzing stratified bilateral correlated data due to its superior performance.
• The score confidence interval provides reliable estimation for the common difference of two proportions.
• Accurate statistical adjustment for both correlation and confounding is crucial for valid medical research inferences.