Modelling binary outcome data

In epidemiology we are most often concerned with deciding how a risk factor is related to disease (or death). If we decide to develop statistical models to represent the relationship between risk factor and disease, it is natural to take the risk factor as the

x

variable and the disease outcome as the

y

variable in the model (Section 9.1). Implicitly, it is the

x

variable that is a potential cause of the

y

variable and not vice versa.