Pannel Data and Fractional Models.

[Juan L. Sobreira]

I have specified my model using fractional probit for a cross section sample (1 year).

The problem is that I wish to check conclusions using a pannel data (8 years); is is possible to implement a "xt" fractional probit using stata?
(Note.- I am working with Stata 14)

[Enrique Pinzon (StataCorp)]

Dear Juan,

The answer to your question is yes, using the population averaged panel probit estimator with robust standard errors, i.e. xtprobit yvar xvars, pa vce(robust) .

[Juan L. Sobreira]

Dear Enrique,

Thanks a lot!!!

[Jeff Wooldridge]

I believe this is the same as xtgee yvar xvars, pa fam(bin) link(probit) vce(robust), which is used in the paper by Papke, L.E. and J.M. Wooldridge (2008), “Panel Data Methods for Fractional Response Variables with an Application to Test Pass Rates," Journal of Econometrics 145, 121-133.


The RE version of xtprobit is not robust with fractional responses (and I believe all of the values of yvar would be automatically changed to zero or one, anyway).

[Jeff Wooldridge]

Ooops, slight edit. One must specify the "unstructured" option, too.

xtgee yvar xvars, pa fam(bin) link(probit) corr(uns) vce(robust)

[Juan L. Sobreira]

Dear Mr. Wooldridge,

Thank you very much for your details