07-23-2012 03:21 PM
I have a question about Zero Inflated Models. We have a very standard survey data set. Our main interest is predicting volunteer hours and "amount of dollars contributed voluntarily."
Both response variables have greater variance than the mean and each has an accompanying variable that asks if the respondent was asked to either "give money" or "give time."
So, we have lots of zeroes (people could have said "no" when asked to volunteer or give money) and the distribution includes excessive zeroes. (e.g. zero hours volunteered and/or zero dollars contributed). We have standard predictor variables, some are ordinal (e.g. age, income, church attendance) others are categorical/class variables. Here are a few questions:
1. Any reccommendations for either inflated negative binomial or zero inflated Poisson models
2. Does anyone know how PROC countreg handles class variables? I am running SAS 9.2 and PROC countreg does not seem to handle the class statement.
3. Does anyone have a brief explanation of how to use the zeromodel command in countreg and what is typically an appropriate variable for the zeromodel portion of the model?
4. If anyone has any examples as code or responses to the above, we would be very grateful.
07-25-2012 10:19 PM
Sorry for my delayed response Dorata_j. I've had a A very busy few days. Thanks for getting back to me so quickly.
PG Stats, Thanks to you as well!! After your note, I looked around and found something telling me I need 9.22 or 9.3 for PROC COUNTREG to take the CLASS statement. Thanks for that tip on upgrading. I have already passed along a request to our admin regards 9.3.
many Thanks all! Gabriel
07-25-2012 07:54 PM
You sometimes refer to Zero Inflated Binomial and sometimes Zero Inflated Negative Binomial. I think you mean the latter, and that led to Dorata_jarosz response (which is right for binomial, but not negative binomial). I don't see what a zero inflated binomial would be.
If you upgrade to 9.3, you have access to PROC FMM which fits ZINB and ZIP models very easily.
As to your statistical question: The Poisson model assumes the conditional variance is equal to the conditional mean. I have never found this assumption to be justified, although I suppose there must be cases where it is.
I am not sure what you mean by "COUNTREG does not seem to handle the CLASS statement". The statement is part of the PROC (see the documentation) and I have used it in the past without problems. Can you post your code and log?