Complete the forward model selection in Example7.9 and compare it with the models selected in Examples7.10 Example

Complete the forward model selection in Example7.9 and compare it with the models selected in Examples7.10

Transcribed Image Text:

Example 7.9 Consider the relationship among gender (g), three-level marital status (m), two-level ed- ucation (e), three-level depression diagnosis (d), and two-level medical burden (c) for the DOS. Since we know that depression is associated with gender, and the association is dif- ferent across the different levels of marital status, we include the three-way interaction of depression, gender, and marital status in the initial log-linear model. We restrict our atten- tion to hierarchical models and thus include all the two-way interactions among the three variables. In the first step of our forward selection process, candidates for inclusion are interactions between the remaining variables, education and medical burden, as well as their interactions with gender, marital status, and depression. The p-values are summarized in the following table. exm Term added exc cxg cxm cxd exd exg p-value 0.0133 0.9505 0.0021 0.0019 0.0148 <0.0001 0.0003 Having the smallest p-value, the interaction between education and gender will first be added. We repeat the procedure based on the revised model with this interaction added. We leave the completion of the procedure as an exercise.