Asymptotic confidence interval construction for proportion ratio based on correlated paired data

Area of Science:

• Biostatistics
• Ophthalmology
• Otolaryngology

Background:

• Paired measurements from paired organs (eyes, ears) in ophthalmological and otolaryngology studies exhibit high intraclass correlation.
• Ignoring this correlation can lead to biased statistical inferences in medical research.
• Accurate confidence intervals are essential for reliable interpretation of proportion ratios from bilateral data.

Purpose of the Study:

• To evaluate different confidence interval (CI) construction methods for proportion ratios with correlated bilateral data.
• To assess the impact of intraclass correlation on CI performance in medical studies.
• To provide a robust methodology for analyzing paired measurements in ophthalmology and otolaryngology.

Main Methods:

• Four confidence interval construction methods were applied: Wald-type, profile likelihood, asymptotic score, and an existing adjusted method.
• Methods were evaluated using Monte Carlo simulations to compare coverage probabilities and interval widths.
• A real dataset from an ophthalmologic study was used for practical illustration.

Main Results:

• The study compared the performance of four different confidence interval methods for proportion ratios.
• Monte Carlo simulations assessed coverage probabilities and widths to identify the most reliable CI construction methods.
• The performance evaluation guides the selection of appropriate statistical methods for correlated bilateral data.

Conclusions:

• Accounting for intraclass correlation is vital when constructing confidence intervals for proportion ratios from bilateral data.
• The evaluated methods offer improved accuracy over ignoring correlation, leading to more reliable inferences.
• The findings are applicable to ophthalmology, otolaryngology, and other fields using paired measurements.