If assumption not met, may
Use polytomous logistic model
Use different ordinal model
Use separate logistic models
Chapter 14: Logistic Regression for
Correlated Data: GEE
If the assumption does not appear to hold, one
option for the researcher would be to use a
polytomous logistic model. Another alternative
would be to select an ordinal model other than
the proportional odds model. A third option
would be to use separate logistic models. The
approach selected should depend on whether
the assumptions underlying the specific model
are met and on the type of inferences the inves-
tigator wishes to make.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 models presented thus far have
assumed that observations are statistically
independent, (i.e., are not correlated). In the
next chapter (Chap. 14), we consider one
approach for dealing with the situation in
which study outcomes are not independent.VII. SUMMARY
üChapter 13: Ordinal Logistic
Regression
This presentation is now complete. We have
described a method of analysis, ordinal
regression, for the situation where the out-
come variable has more than two ordered
categories. The proportional odds model was
described in detail. This may be used if the
proportional odds assumption is reasonable.Presentation: VI. Ordinal vs. Multiple Standard Logistic Regressions 481