Analysis of symmetric cross-classifications

摘要: Analysis of data involving symmetrically classified responses involves log-linear models with nonstandard parmetrization. These can be specified in an extended version symbolic notation that is available GLIM, so suitable definition qualitative and quantitative terms the recognition nesting relations among lead to a unified treatment large number previously proposed methods. This approach may (i) comparison symmetric tables different populations (ii) modelling effects cross-classified factors on response distribution from exponential family. Illustrative applications involve eye vision men women dependence fertility father’s respondent’s social class.