ABSTRACT
The chapter is concerned with classical models for tables of counts, including contingency tables. The models are presented as examples of curved or regular exponential families and thus illustrate the developments in previous chapters. Classical test statistics are derived as likelihood ratio statistics and Wald statistics. Models investigated include models for testing goodness-of-fit, genetic models such as the Hardy-Weinberg law, comparison of multinomial distributions, independence of classification criteria, the multiplicative Poisson model, also including background exposure. Measures of association such as relative risk, odds-ratios, etc. are described together with methods for constructing confidence intervals for these. Monte Carlo methods for the analysis are briefly described, including so called exact tests. All concepts are illustrated with analyses of small, real datasets. Finally, test for association of ordinal data are briefly described.
