Richard Williams <[email protected]> wrote,
> I wonder if any user-written post-estimation commands will get
> zapped...factor variables especially sound like they might zap
> something that was looking for the actual variables used, but maybe not.
Richard is right in suspecting there are changes in how estimation
results are stored. Under version control, however, results are
stored just as they have always been and, usually, there's no issue
anyway, so you don't even have to run under version control. To use
the commands with factor variables, however, they will have be updated.
Here are the details:
1. Existing (as of this date) user-written post-estimation
commands (henceforth "older commands") will not work when
factor variables are specified in the model. Older commands
will have to be updated so that they understand factor
variables.
2. Older commands mostly work fine when factor variables are not
specified in the model, even without specifying -version 10-
in front of them.
Stata 11 marks dropped variables in estimation commands just
as we have always marked them in -regress-. The change is
that we are now using the -regress- method for all estimation
commands. Thus, it is when variables were dropped during
estimation that problems can arise.
If the older command worked with after -regress-, it will work
without version control. If the older command did not work,
if it was not intended to work after -regress-, it might work
if there are dropped variables, and it might not. It all
depends on how the programmer wrote it. If the older command
does not work in the presence of dropped variables, it usually
will still work if there are no dropped variables.
3. No matter what, if you run both the estimation command and the
older post-estimation command under version control, results
will be just as they were in Stata 10.
-- Bill
[email protected]
*
* 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/
