## ABSTRACT

In Chapter 4 we got an impression of the versatility of the linear normal model. From a theoretical point of view, this class of models is relatively simple and this may also be said about the calculations required by the users of those models provided that they can use a statistical computer package, for example, the procedure PROC GLM in SAS. However, as the name of this class of models indicates, they can be applied only in connection with normally distributed data. It is therefore natural to investigate whether the relatively simple structure in linear normal models may be generalized to data that are not normally distributed, such as, for instance, discrete data. In this chapter we give a brief introduction to a generalization of the linear normal model, which in literature is known as the generalized linear model. Some statistical computer packages contain procedures that can perform the calculations in this class of models. This also applies to SAS where the calculations may be carried out by means of the procedure PROC GENMOD.