Dear Statalisters,
This may be a bit too un-stata-related for the statalist, but please, let me give it a try.
I am using a model which has elicited some interest from several people from the list recently: the semiparametric Tobit Fixed Effect model by Bo Honore ("Trimmed LAD and Least Squares Estimation of Truncated and Censored Regression Models with Fixed Effects." Econometrica, May, 1992, Vol. 60, No 3, pp. 533-565.)
I am using a two-year panel data set and in the final specification, I want to control for _time-variant_ regional effects by introducing a set of regional dummies interacted by a time dummy for year 2. Of course, one of the dummies is dropped.
In a linear setting, changing the reference category only affects the constant term and the coefficients on the regional dummies, but leaves the remaining coefficients unchanged. I am very puzzled by the following results: in Tobit FE, when I choose a different reference category, the coefficients on the other (non regional) variables change as well.
That is: using TOBIT FIXED EFFECTS (and say, there are three regions)
Y=a0+ a1X1 +a2X2 + ... + region2*T2 + region3*T2 --> dropping region 1
Y=a0+ a1X1 +a2X2 + ... + region1*T2 + region3*T2 --> dropping region 2
and the coefficients a2 changes a lot between the two specifications.
Would anyone have a sense of what is happening here? I am contacting people from my university to clarify this matter, but it looks like it is quite specialised and I would greatly appreciate any help from the list.
Many thanks to all of you for making the statalist such a helpful community, and my very best wishes for the New Year.
Thi Minh
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
