Confidence interval construction for a difference between two dependent intraclass correlation coefficients

Abstract

Inferences for the difference between two dependent intraclass correlation coefficients (ICCs) may arise in studies in which a sample of subjects are each assessed several times with a new device and a standard. The ICC estimates for the two devices may then be compared using a test of significance. However, a confidence interval for a difference between two ICCs is more informative since it combines point estimation and hypothesis testing into a single inference statement. We propose a procedure that uses confidence limits for a single ICC to recover variance estimates needed to set confidence limits for the difference. An advantage of this approach is that it provides a confidence interval that reflects the underlying sampling distribution. Simulation results show that this method performs very well in terms of overall coverage percentage and tail errors. Two data sets are used to illustrate this procedure.

Confidence Interval Estimation for a Difference between Two Dependent Intraclass Correlation Coefficients with Variable Class Sizes

“…The results can be seen as a direct extension of Ramasundarahettige, Donner and Zou (2009). Previous research has largely focused on hypothesis testing of equality (Donner, Koval and Bull, 1984), but confidence intervals are more informative as they indicate statistical significance as well as the magnitude of the difference, providing basis for judging practical importance.…”

“…A confidence interval for the difference between the sister-sister and brother-brother ICCs would be more informative than a P-value resulted from hypothesis testing. Ramasundarahettige, Donner and Zou (2009) recently applied a general confidence interval procedure for a difference to the case of dependent ICCs in the context of reliability studies having constant class sizes. Since the basic idea is to use confidence limits for single ICCs to recover variance estimates needed for a difference, we refer to the method as the MOVER, method of variance estimates recovery (Zou and Donner, 2008;Zou, 2008).…”

“…Application of the MOVER resulted in four confidence interval procedures for a difference between two dependent ICCs. As in Ramasundarahettige, Donner and Zou (2009), we take into account dependency between two ICCs using a result from Elston (1975).…”

A Confidence Interval Approach for Comparative Studies Involving Binary Outcomes in Paired Organs

“…For this purpose, we first review the add-and-square, also known as the method of MOVER, confidence interval construction approach proposed by Newcombe (1998b) and Zou and Donner (2008) for proportion difference. The MOVER has been widely adopted by Zou and his fellows in epidemiological studies (Zou, 2008), psychological studies (Zou, 2007), environmental studies , medical studies (Zou and Donner, 2008;Ramasundarahettige et al, 2009), and binary outcome studies . Consider the point estimateπ i = (m 1i + 2m 2i )/(2m +i ) for π i (i = 0, 1).…”

Comparison and Reproducibility of Nuchal Translucency Measurements Using Two-Dimensional and Volume Nuchal Translucency Ultrasound: A Prospective Study

“…Comparison of two independent ICCs was based on a confidence interval around their difference [13]. Continuous variables were reported as means ± SD.…”