6 Ordered Regression Models
The sample size does not change across the cutpoint equations, which is a feature of
cumulative models that is not shared by stage or adjacent models. For example, if there
are three cutpoint equations, then every respondent is coded 0 or 1 for y
m
in all three
binary cutpoint equations. A respondent’s value for y
m
will vary across cutpoint equa-
tions, but the key point is that every respondent is included in each equation in the
cumulative model. As a result, the binary equations are not independent, which has
important implications for model estimation. We discuss this issue in more detail in
Chapter2. If the ordinal outcome is based on a sequence of stages with a logical start-
ing point, then the conditional probab ...