Dear Statalist users,
I am trying to fit some advanced models on censored (at zero) data, and I
wonder if anybody could point out an approach to do this in Stata (or in
other software). I will refer to this models as Tobit because that?s the
type of model I fitted so far, but I does not have to be strictly a Tobit,
as long as it is a proper model for censored data. However, I did not use
the Heckman two-steps approach because, for identification, it requires that
some variables affect the selection process but not the value of the outcome
if the variable is observed. This is not the case in my applied problem. The
model I would like to fit are:
1) Tobit with a parametric specification for the variance to control for
possible heteroskedasticity, for example in which the variance is a linear
function of some exogenous variables. Or at least including weights to
reduce the effect of heteroskedasticity;
2) Tobit the same as 1) but for static panel data (i.e. with a random effect
component);
3) Tobit for static panel data (random effect) with instrumental variables
to control for continuous endogenous predictors;
4) Multivariate Tobit for panel data;
I know I am asking how to do quite demanding analysis, and I would be really
grateful is somebody would be able to point out some possible solutions,
also with other software. Again, these models does not necessary have to be
?Tobit?, as long as able to deal with censored observations.
All the very best,
Carlo
**************************************************
Carlo Fezzi
Senior Research Associate
Centre for Social Research on the Global Environment (CSERGE)
Department of Environmental Sciences
University of East Anglia
Norwich (UK) NR2 7TJ
Telephone: +44(0)1603 591385
Fax: +44(0)1603 593739
*************************************************
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/