A generalized concordance correlation coefficient for continuous and categorical data

Abstract

This paper discusses a generalized version of the concordance correlation coefficient for agreement data. The concordance correlation coefficient evaluates the accuracy and precision between two measures, and is based on the expected value of the squared function of distance. We have generalized this coefficient by applying alternative functions of distance to produce more robust versions of the concordance correlation coefficient. In this paper we extend the application of this class of estimators to categorical data as well, and demonstrate similarities to the kappa and weighted kappa statistics. We also introduce a stratified concordance correlation coefficient which adjusts for explanatory factors, and an extended concordance correlation coefficient which measures agreement among more than two responses. With these extensions, the generalized concordance correlation coefficient provides a unifying approach to assessing agreement among two or more measures that are either continuous or categorical in scale.