METHODS FOR DETECTING AND ASSESSING THE INTERRATER BIAS

Abstract

In this paper we present an analysis of the different methods commonly used to detect and assess interrater bias from two basic perspectives: classical approach based on descriptive-type criteria and loglinear model approach. We show that all these procedures are not satisfactory for the detection and measurement of observer bias in a univoque way due to contradictory results. We propose some new alternatives to develop which allow the correct separation of bias and measurement error.