Logistic Regression: A Self-learning Text, Third Edition (Statistics in the Health Sciences)

Which model? Requiresstrategy

3. Computing the Odds Ratio

It should not be surprising to see different

values for corresponding coefficients as the

two models give a different description of the

underlying relationship among the variables.

To decide which of these models, or maybe

what other model, is more appropriate for

this data, we need to use astrategyfor model

selection that includes carrying out tests of

significance. A discussion of such a strategy is

beyond the scope of this presentation but is

described elsewhere (see Chaps. 6 and 7).

This presentation is now complete. We have

described important special cases of the logis-

tic model, namely, models for

We suggest that you review the material cov-

ered here by reading the detailed outline that

follows. Then do the practice exercises and

test.

All of the special cases in this presentation

involved a (0, 1) exposure variable. In the next

chapter, we consider how the odds ratio for-

mula is modified for other codings of single

exposures and also examine several exposure

variables in the same model, controlling for

potential confounders and effect modifiers.

SUMMARY

1. Introduction
   3 2. Important Special Cases

 simple analysis

 interaction assessment involving two

variables

 assessment of potential confounding and

interaction effects of several covariates

64 2. Important Special Cases of the Logistic Model