xtprobit - using lags of the dependent variable

[Alex Jeffries]

Hello there,

I am working with a panel dataset with lots of dummys, and want to run a probit to work out the probability in a given period of taking out insurance. As my independent variables I have dummies such as has there been a fire before, was the income in the last period above a certain value etc. What I would also like to do is have lags of the insurance dummy, so whether insurance was taken out last period, and the period before that.

Edit: Solved that.
My issue now is after running the probit, I want to work out the marginal effects, but when I do
margins, dydx(*) atmeans
I get the error
default prediction is a function of possibly stochastic quantities other than e(b),
what does this mean?

[Carlo Lazzaro]

Alex:
I would take a look at: http://www.ats.ucla.edu/stat/stata/f...logit_prob.htm

[Alex Jeffries]

Alex:
I would take a look at: http://www.ats.ucla.edu/stat/stata/f...logit_prob.htm

Hi Carlo, thanks for replying. I had come across that resource yesterday and tried to follow the steps there already, but I am still having issues. I've read through the stata help files and remarks but I am making little progress.

From that link they use
margins female, at(read=(30(10)70)) predict(mu fixedonly) vsquish
If I try margins firedummy, predict(mu fixedonly) vsquish, I get that option mu is not allowed. I can't seem to figure out what the mu is, as there is no mention of it in that page, (and I also don't see vsquish in the help files either). I have never seen binary choice models before so I am very confused.

[Carlo Lazzaro]

Alex:
mu is the predicted mean of the fixed part of the model (as explained in the web page);
-vsquish- is described in -estimation options-.