Antwort: st: Heckprob error: outcome = y_select > 0 predicts data perfectly

[Justina Fischer <JFischer@diw.de>]

Hi

this is eactly your problem. If they are omitted form the regression sample of the main rgresion, the first stage is pefectly predicted by your 'non-response' variable.

Try to gone the missings a certain category of reply (0, 99, whatever) and it should work

Justina
-----owner-statalist@hsphsun2.harvard.edu schrieb: -----

An: statalist@hsphsun2.harvard.edu
Von: Dave Francis <dafrancis11@yahoo.com>
Gesendet von: owner-statalist@hsphsun2.harvard.edu
Datum: 27.01.2011 08:58PM
Thema: st: Heckprob error: outcome = y_select > 0 predicts data perfectly

Hi,

I am trying to a heckprob model on the disappearance of firms. Basically, I have
a selection equation where firms are coded 0 if they existed in the first stage
of my panel but not the second statge. The firms are coded 1 if they are in
existence in both periods of the panel. Secondly, I have a binary outcome stage
equation, which is equal to 1 if a firm answers a specific question; 0 if it
does not.


I'm getting this error: outcome = y_select > 0 predicts data perfectly, where
y_select is the dependent variable for my selection equation. Thoughts? Clearly
since firms that are not "selected" are those that are no longer in existence,
they cannot answer and so they are coded as missing for y_outcome, the DV of the
outcome equation.

Thoughts?

Many thanks

Here is a tab of the two y_* vars

. tab y_outcome y_select, m
|         y_select
y_outcome  |   0          1 |     Total
-----------+----------------------+----------

0 |         0         95 |        95
1 |         0        178 |       178
. |        70          0 |        70

-----------+----------------------+----------
Total |        70        273 |       343




*
*   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/