ABSTRACT
In the last two chapters, we discussed how to make inference about association between two variables with stratification (sets of contingency table) and without stratification (a single contingency table) by a third categorical variable. In such analyses, our primary interest lies in whether the two variables are associated as well as the direction of association, with stratification used to control for the effect of the categorical confounder. If there are many confounders or the confounding variable is continuous, the methods discussed in Chapter 3 for stratified tables may not work well or do not work at all. Furthermore, in many studies, we may want to know more about the relationship between the variables. Specifically, we want to know the amount of change in one variable per unit change of the other variable. The methods discussed in the last two chapters lack such specificity, and regression models, which are the focus of this chapter, come to rescue.
