Date of Award

2008

Degree Type

Thesis

Degree Name

Master of Science

Program

Epidemiology and Biostatistics

Abstract

The intraclass correlation coefficient (ICC), an index of similarity, plays an important role in a wide range of disciplines, for example in the assessment of instrument reliability. In this case, the study design may involve recruiting a sample of subjects each of whom are assessed severe^ times with a new device and the standard. The ICC estimates for the two devices may then be compared using a test of hypothesis. However it is well known that conclusions drawn from hypothesis testing are confounded by sample size, i.e., a significant p-value can result from a sufficiently large sample size. In such cases, 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. The sampling distribution for the ICC is well known to be left-skewed and thus confidence limits are usually constructed using Fisher’s Z-transformation or the F- distribution. Unfortunately, such an approach is not applicable to a difference between two ICCs. The remaining alternative is to apply a simple asymptotic approach, i.e., point estimate plus/minus normal quantile multiplied by the estimate of standard error. However this method is known to perform poorly because it ignores the features of the underlying sampling distribution. In this thesis I develop a confidence interval procedure using the method of variance estimate recovery (MOVER). Specifically, the variance estimates required for the upper and lower limits of a difference are iii recovered from those obtained for separate ICCs. An advantage of this approach is that it provides a confidence interval that reflects the underlying sampling distribution. Simulation results show that the MOVER method performs very well in terms of overall coverage percentage and tail errors. Two data sets are used to illustrate this procedure.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.