The models presented later describe how both the marginal distributions of the responses, and their association, vary with explanatory variables. Here, we preview those data with a fourfold display focused on just the association (odds ratio) between the two binary response variables, Hypertension and Urea.
Figure 3.XX shows the fourfold displays for the 15 combinations of mother's social class and smoking, with non-smokers in the top row, and heaviest smokers in the bottom row. We see that, with one exception (a cell with small total frequency), the association between presence of hypertension and urea is positive, and roughly of the same magnitude over the categories of class, particularly for non- and moderate-smokers. In Example 7.15 we plot the log odds ratio directly (Figure 7.32), as well as the log odds for prevalence of these two symptoms.