Hi all,
I'm implementing a two-part model, using tpm (from SSC; first part probit and then GLM with Poisson or IG family with a log link). I am trying to get the percent change in the outcome due to a change in a binary predictor (when all other variables are held at their average value).
I have referred to the help files but still have a couple of questions about the postestimation commands available:
I'm implementing a two-part model, using tpm (from SSC; first part probit and then GLM with Poisson or IG family with a log link). I am trying to get the percent change in the outcome due to a change in a binary predictor (when all other variables are held at their average value).
I have referred to the help files but still have a couple of questions about the postestimation commands available:
- Does the predict command report the dependent variable when the covariates in the model are at the average value or the reference value?
- After using tpm, how does the predict command compute the outcome for each observation? Does it first compute the value when the binary predictor of interest==0, then compute the value when the predictor==1, then compute the difference, then take the average of the differences?