Confidence intervals for proportion ratios of stratified correlated bilateral data

• Published in: Journal of biopharmaceutical statistics 2019-01, Vol.29 (1), p.203-225
• Main Authors: Zhuang, Tingting, Tian, Guo-Liang, Ma, Chang-Xing
• Format: Article
• Language: English
• Subjects: Anti-Bacterial Agents - therapeutic use; Biostatistics - methods; Collagen - therapeutic use; Computer Simulation; Confidence interval; Confidence Intervals; Data Interpretation, Statistical; equal correlation coefficients model; Humans; Likelihood Functions; Monte Carlo Method; Otitis Media with Effusion - drug therapy; Otitis Media with Effusion - microbiology; proportion ratios; Randomized Controlled Trials as Topic - statistics & numerical data; Research Design - statistics & numerical data; Scleroderma, Diffuse - drug therapy; Scleroderma, Diffuse - pathology; stratified bilateral data
• ISSN: 1054-3406; 1520-5711
• DOI: 10.1080/10543406.2018.1489405

In stratified bilateral studies, responses from two paired body parts are correlated. Confidence intervals (CIs), which reveal various features of the data, should take the correlations into account. In this article, five CI methods (sample-size weighted naïve Maximum likelihood estimation (MLE)-based Wald-type CI, complete MLE-based Wald-type CI, profile likelihood CI, MLE-based score CI and pooled MLE-based Wald-type CI) are derived for proportion ratios under the assumption of equal correlation coefficient within each stratum. Monte Carlo simulation shows that the complete MLE-based Wald-type CI approach generally produces the shortest mean interval width and satisfactory empirical coverage probability with close form solution; while the profile likelihood CI and the MLE-based score CI provide preferred ratio of non coverage probability and are more symmetric. Two real examples are used to demonstrate the performance of the proposed methods.