Generalized Linear Models

squares for balanced mixed models. I agree with the author that this approach is easier and more understandable than the usual expected mean square rules using tables of subscripts with live and dead indices (Kuehl, 2000, section 7.6). Rules are also given for calculating variances of means and pairwise differences in means for balanced mixed models. There is also a brief description of how to use statistical software to Ž nd expected mean squares for unbalanced mixed models. The material on the analysis of variance is supplemented with a short presentation on linear models for Ž xed effects in the Appendix.