This monograph deals with a class of statistical models that generalizes classical linear models to include many other models that have been found useful in statistical analysis. These other models include log-linear models for the analysis of data in the form of counts, probit and logit models for data in the form of proportions (ratios ofcounts), and models for continuous data with constant proportional standard error. In addition, important types of models for survival data are covered by the class. An important aspect of the generalization is the presence in all the models of a linear predictor based on a linear combination of explanatory or stimulus variables. The variables may be continuous or categorical (or indeed a mixture of the two), and the existence of a linear predictor means that the concepts of classical regression and analysis-of-variance models, insofar as they refer to the estim­ ation of parameters in a linear predictor, carry across directly to the wider class of model. In particular, the ideas underlying facto­ rial models, including those of additivity, interaction, polynomial contrasts, aliasing, etc., all appear in the wider context. Generalized linear models have a common algorithm for the est­ imation of parameters by maximum likelihood; this uses weighted least squares with an adjusted dependent variate, and does not require preliminary guesses to be made of the parameter values.