ABSTRACT
The previous two chapters introduced the logistic regression model and provided the necessary tools to fit such models to data, including that arising from case-control designs. Chapters 13.2 and 13.4 contrasted results from this approach with those previously obtained for categorical exposures. Although we briefly tackled the modeling of an interval-scaled exposure (body weight) in Table 13.3, we so far seem to have introduced the heavy machinery (of regression) merely to do a job effectively performed by hammers and screwdrivers (simple categorical methods of Chapters 7 to 11). What’s the point? Are we just trying to complicate matters for effect? Well, of course not. While the previous chapters introduced logistic regression, linking these simple cases with our previous work, we now want to take advantage of the generality of regression methods so that we can achieve the goals we promised, in particular, simultaneous adjustment of several risk factors and confounders using a modest number of degrees of freedom.
