|
Activity Number:
|
68
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 2, 2009 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
| Abstract - #303600 |
|
Title:
|
A Note on the Estimation of Nonparametric Correlation Coefficients for Clustered Data
|
|
Author(s):
|
Allison M. Deal*+ and Haitao Chu and Bert O'Neil
|
|
Companies:
|
The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill
|
|
Address:
|
450 West Drive, Lineberger Cancer Center, Chapel Hill, NC, 27599,
|
|
Keywords:
|
nonparametric correlation coefficient ; clustered data ; percentile method ; bootstrapping methods ; iterated method
|
|
Abstract:
|
In many clinical studies, nonparametric correlation coefficients are used to summarize the association between two continuous variables. When the paired random variables are repeatedly measured per individual, the independence assumption is usually not satisfied. Thus, it may be problematic to use standard nonparametric methods ignoring clustering to construct confidence intervals. We investigate the performance of percentile and iterated bootstrapping methods on this issue through simulations and a real data example. Simulations revealed increasingly inflated Type I error rates as the intraclass correlation coefficient and the number of observations per person increased if clustering is ignored; thus, it is recommended to compute p-values. Furthermore, simulations revealed that an iterated bootstrapping method produced confidence intervals with 95% coverage probability.
|
