The measurement of interobserver agreement based on categorical scales

Abstract

The Kappa statistic is used to measure the interobserver similarity based on categorical scales. The cases of two or more observers with two or more rating categories are considered. Allowance is made for the attachment of disagreement weights, based on rational or clinical grounds, to different rating categories. Tests of hypotheses about the conditions Kappa = 0 and Kappa > 0 are conducted.