From: "Nick Cox" <[email protected]>
Reply-To: [email protected]
To: <[email protected]>
Subject: st: RE: bugs using _all
Date: Sun, 10 Jul 2005 22:30:38 +0100
The Statalist FAQ advises this:
--------------------
Don't say "Is this a bug?". Almost all the time, it is not!
--------------------
However, you have, I guess been bitten by something
designed to protect you.
When you type -logit-, you fire up a wrapper program
which, among other things, creates a temporary variable
recording the sort order, so that your data can be
left in their current sort order when -logit- is done.
This temporary variable is thus part of _all when
-logit- passes the ball. Evidently it is true that
your data order are in the sort order of your response
variable, so -logit- bails out for the reason given.
This is fixable by StataCorp, although I am not
sure how keen they will be to do it. The code to
fix it is embedded within -egen.ado-, for which
the fix is arguably more important.
The easiest fix is just not to do this, in the way
that you have demonstrated. In most circumstances,
firing all the variables at a binary response is
likely to be poor science. I guess in your case
it just seemed a neat short-cut.
I can't comment on your other cases not documented
here.
Nick
[email protected]
Jun Xu
> Could be that I missed something, but it might be a bug. I found that
> sometimes, I will get weird results or simply refusal to
> estimate a simple
> logit model. For example (lfp is the first variable in the list),
>
> . logit _all
>
> outcome = __000000 > 325 predicts data perfectly
> r(2000);
>
> . logit lfp-inc
*
* 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/
_________________________________________________________________
