ABSTRACT
An important characteristic of generalized linear models is that they assume independent observations. The simplest instance of a model excluded by this criterion is the standard linear model for the splitplot design, which has two error terms, one for between-whole-plot variance and one for within-whole-plot variance. For instance autoregressive models can easily be fitted using programmes designed expressly for ordinary linear models. In generalized linear models, additivity is, correctly, postulated as a property of the expected responses. Having selected a particular model, it is required to estimate the parameters and to assess the precision of the estimates. In the case of generalized linear models, estimation proceeds by defining a measure of goodness of fit between the observed data and the fitted values generated by the model. Residuals can be used to explore the adequacy of fit of a model, in respect of choice of variance function, link function and terms in the linear predictor.
