105Nonparallel Models
4.5.1 Guidelines for Choosing a Parallel, Partial, or Nonparallel
Ordered Regression Model
In the previous chapter, we suggested that researchers should consider two main factors
when choosing a parallel or partial ordered regression model. Parallel models are pre-
ferred if there is no ambiguity in the ordering of outcome categories, and the parallel
regression assumption is a reasonable assumption for each independent variable. This
applies when we consider nonparallel models as well.
If the parallel regression assumption is too restrictive for one or more variables in the
model, then the researcher must decide whether to relax the assumption for the entire
model or on a variable-specic basis. Nonparallel models are ...