I am using svymlogit to perform a multinomial logistic regression with survey data. A couple of
questions:
-Is there a way to assign a unit to a continuous predictor? For example, in SAS, I could assign
units of 10,000 to my income variable, so that the coefficient and odds ratio will correspond to
increases of $10,000 rather than increases of $1, which is not very meaningful.
-svymlogit gives me an overall F value with df based on strata and psu's. Is this F value the same
as -2LL? If not, is there some way I can get that value?
- Is there a way to test the whole set of possible interactions - for example, in a nested fashion
- before including each individual interaction? And should I be using inteff? Why can't I just
enter the interaction terms into the svymlogit model? What is the logic behind inteff?
