
90 Ordered Regression Models
some extent for age and age-squared. On average, the effects of age and age-squared on
the predicted probabilities for each level of health are more modest than the odds/hazard
ratios would suggest.
The average effects of sex and race/ethnicity on the predicted probabilities are not sta-
tistically signicant, which is what we would expect based on the raw coefcients and
odds/hazard ratios. Additionally, marriage only has a signicant effect on the probability
of poor health, which is consistent with the previous results. On average, being married
is associated with a 0.03 decrease in the predicted probability of poor ...