A WEIGHTED CONCORDANCE CORRELATION-COEFFICIENT FOR REPEATED MEASUREMENT DESIGNS

The need to quantify agreement between two raters or two methods of me asuring a response often arises in research. Kappa statistics (unweigh ted and weighted) are appropriate when the data are nominal or ordinal , whereas the concordance correlation coefficient is more appropriate when the data are measured on a continuous scale. We develop weighted product-moment and concordance correlation coefficients which are appl icable for repeated measurements study designs. We consider two distin ct situations in which the repeated measurements are paired or unpaire d over time. We illustrate the methodology with examples comparing (1) two assays for measuring serum cholesterol, (2) two estimates of diet ary intake, from a food frequency questionnaire and dietary recalls, a nd (3) two measurements of percentage body fat, from skinfold calipers and dual energy x-ray absorptiometry.