Introduction This chapter continues the previous chapter (Chap. 6) that
gives general guidelines for a strategy for determining a
best model using a logistic regression procedure. The focus
of this chapter is the interaction and confounding assess-
ment stages of the model building strategy.
We begin by reviewing the previously recommended
(Chap. 6) three-stage strategy. The initial model is required
to be hierarchically well formulated. In carrying out this
strategy, statistical testing is allowed for assessing interac-
tion terms but is not allowed for assessing confounding.For any interaction term found significant, a hierarchy
principle is required to identify lower order variables that
must remain in all further models considered. A flow dia-
gram is provided to describe the steps involved in interac-
tion assessment. Methods for significance testing for
interaction terms are provided.Confounding assessment is then described, first when
there is no interaction, and then when there is interaction –
the latter often being difficult to accomplish in practice.Finally, an application of the use of the entire recom-
mended strategy is described, and a summary of the strat-
egy is given.Abbreviated
Outline
The outline below gives the user a preview of the material
to be covered in this chapter. A detailed outline for review
purposes follows the presentation.I. Overview (pages 206–207)
II. Interaction assessment stage (pages 207–210)
III. Confounding and precision assessment when no
interaction (pages 211–215)
IV. Confounding assessment with interaction (pages
215–223)
V. The Evans County example continued (pages
223–230)
VI. Summary (pages 231–232)204 7. Modeling Strategy for Assessing Interaction and Confounding