02-10-2017 12:22 PM
I am running a multinomial logistic regression with proc logistic. The dependent variable is nomial, 0, 1, 2, 3 four categories, the model has four predictors, see Age, Sex, Education and Race, I could use dependent variable=0 as reference, so I could get 1 vs 0, 2 vs 0 and 3 vs 0, but I also want to know how it looks like such as 1 vs 2, 2 vs 3 and 1 vs 3, it is possible to those results by only runing model once with 0 as reference? Or I have to run model several times with dependent categories as reference? if not for proc logistic, how about proc genmod or other procedures? Thank you.
02-14-2017 11:39 AM
Check out the ODDSRATIO statement.with a DIFF option. The default is DIFF=ALL which would give you the comparisons you are looking for.