--- Rachel <[email protected]> wrote:
> I have a panel dataset with t=3. I'm doing a simple probit equation
> for a binary dependent variable in each of the time periods. All
> independent variables are time invariant.
>
> For t=1, the predicted classifications for the probit (using -estat
> clas-) product a lot of false negatives.
>
> What could be causing the trouble? Would including a time dummy
> help, or do I have to use fixed effects?
Rachel:
If your predictors aren't very informative and there are more negatives
than positives, then -probit- will give the best prediction possible:
everybody a negative. This would thus result in lots of false
negatives. So, it means that your predictors don't predict very well.
This is by no means unique for -probit-: within the social sciences it
is quite common to get a proportion of explained variance of less then
10% when using -regress-, i.e. more than 90% of the variance in the
"explained" variable remains unexplained. Some suggested to call the
Social Sciences the hard sciences because it is so hard to get a
reasonable model. Depending on your discipline you might just have to
learn to live with that many false negatives.
You are ignoring the panel structure of your data. Not doing that, for
instance by using -xtprobit- instead of -probit-, might help.
Hope this helps,
Maarten
-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands
visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434
+31 20 5986715
http://home.fsw.vu.nl/m.buis/
-----------------------------------------
Send instant messages to your online friends http://uk.messenger.yahoo.com
*
* 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/
