
124 Ordered Regression Models
very similar. The differences for excellent health are somewhat larger, but the overall age
pattern is approximately the same in both models. Yet, the differences for good health are
more important and may warrant relaxing the parallel regression assumption. There is an
increase in the predicted probability of good health after the age of 60 that is only detect-
able by relaxing the parallel regression assumption.
The differences in the precision of the estimates are more substantial. There are similar
patterns in the condence intervals for fair and excellent health, but the patterns diverge
for poor and good health. ...