I am currently using the ordered probit model for rate of alcohol consumption Y=0 (Not drinking), Y=1 (Light drinker) and Y=2 (heavy drinker). Using "margins, dydx(*)" command on Stata i got the marginal effects of the independent variables for Y=0 (Not drinking), Y=1 (Light drinker) and Y=2 (heavy drinker). Is it logical to report the marginal effect of Y=0 or just Y=1 and Y=2?