Quantitative methods for the analysis of observer agreement: toward a unifying model
This paper reviews a variety of recently proposed models for the analysis of observer agreement. The discussion includes latent class and latent trait models, signal detection theory models, item response and Rasch models, association and quasi-symmetry models, and methods based on correspondence analysis and multidimensional scaling. Advantages and limitations of each approach are discussed and particular attention is paid to relationships among the approaches. Relationships between these approaches and the kappa coefficient are also observed. The paper concludes with a discussion of the general requirements for a general model for the analysis of observer agreement and suggests possible directions for further research in this area.