[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: Random effects probit models (xtprobit)--normalisation, normaly tests and storing residuals
I have estimated a random effects probit model using the default xtprobit
(aghermite) command in Stata 9.0. I employed Wooldridge's (2005),Journal of
Applied Econometrics, solution to the initial conditions problem and
therefore the resulting "rho" produced by Stata is :
Sigma squared ai/(1+ sigma squared ai)
i.e if we have a composite error term : ai+uit, uit has a conditional
standard normal distribution.
Is it possible to use another normalisation isntead? I.e sigma squared ai+
sigma squared uit=1??? (Vella and Verbeek 1998, Journal of Applied
Econometrics) use such a normalisation.
How can this be implemented?
And then how can someone conduct a conditional moment test of normality for
both (ai, uit)?
Or just for (ai)?
In my understanding Stata does not have an available post estimation command
( following xtprobit) for saving the error terms ( e.g in linear models one
can use predict r, residuals). The only options available are xb ( linear
prediction ) and stdp(standard error of the linear prediction) but to me
(and I might well be wrong) neither of these seem to be relevant.
Would be most grateful if anyone could help!
* For searches and help try: