Log-linear models for association and agreement in stratified square contingency tables (Q1966034)

Simple and interpretable models are introduced for classifications of nominal and ordinal scales which restrict only the symmetrical association and leaving the asymmetrical part arbitrary. The corresponding models for the asymmetric part of the association are discussed. All considered models are log-linear models which can be fitted by standard software. The methods are applied to an original data set on the classification of uterine cancer and social mobility tables.