[Introduction of the general linear models]

General linear models can be considered the paradigm of all models used in clinical epidemiology. In these models, the independent variables combine in linear fashion to predict the values of the variable response. Since no model predicts the variable response perfectly, an error term is incorporated into the model to acknowledge what remains to be explained after getting a fit to the data. When this error term is normally distributed with constant variance, the linear models are reasonably appropriate to describe the input/output relationship of interest.