Quadratic Estimating Equations Based Confidence Interval of the Intraclass Correlation Coefficient for Binary Outcome Data

• Activity Number: 71
• Type: Contributed
• Date/Time: Sunday, July 31, 2011 : 4:00 PM to 5:50 PM
• Sponsor: Biometrics Section
• Author(s): Krishna K. Saha*+
• Companies: Central Connecticut State University
• Keywords: Intraclass correlation parameter ; Binary data ; Confidence interval ; Hybrid profile variance ; Profile likelihood

Abstract

This paper investigates four procedures for computing the asymptotic confidence intervals of the intraclass correlation coefficient (ICC) for clustered binary outcome data based on the parametric and semiparametric models. We derive closed -form asymptotic and sandwich variance formulae for four point estimators obtained using the parametric and semiparametric models. We also introduce an confidence interval for the ICC using a modified Wald test proposed by Rao and Mukerjee (1997) based on the hybrid profile variance (HPV) for the QEE. We then compare the proposed approach with asymptotic approaches through simulations in terms of coverage and interval length. The results indicate that the asymptotic approaches have severe under-coverage, whereas the proposed HPV approach perform quite well by providing coverage very close to nominal levels over a wide range of parameter combinations. An example to toxicological binary outcome data is analyzed.