Comparison of Concordance Correlation Coefficient Estimating Approaches with Skewed Data

Abstract

The concordance correlation coefficient (CCC) is an index that assesses the agreement between continuous measures made by different observers. At least four methods are used to estimate the CCC: two (Lin's method, Variance Components) which are defined on the basis that data are normally distributed, and the two others (U-statistics, GEE) which do not assume any particular distribution of the data. Here the four methods are compared with skewed data from a model in which the subject means follow a log-normal distribution while the within-subject variability is assumed to be normally distributed. An example of alcohol consumption is considered and a simulation study is performed.

Key Words:

• Concordance correlation coefficient
• Generalized estimating equations
• Linear mixed model
• Log-normal distribution
• Skewed data
• U-statistics
• Variance components

ACKNOWLEDGMENT

We are indebted to Dr. Ricard Tresserres (Public Health Department, Government of Catalonia) for kindly delivery the alcohol consumption data. We would like to thank to Huiman Barnhart for providing the SAS macro that estimates the CCC using the GEE approach.

Notes

^aAssuming asymptotic normality.