I performed your first suggestions - with e(sample) - and the results appear logical. I like this method better than imputing or using the forced option ( read about this and it seems risky if one -like myself- is not skilled at this method.
I will try out the "gen touse" option and compare the lrtest statistic with my present outcome and determine if there are large differences.
I really appreciate your help. I was trying to figure this out for a few days. I didn't know about the "e(sample)" option.
Thanks a loads...Semilla
-----Original Message-----
From: Richard Williams [mailto:[email protected]]
Sent: Fri 3/5/2004 10:52 PM
To: [email protected]
Cc:
Subject: RE: st: likelihood ratio test error message
At 09:28 PM 3/5/2004 -0600, Rivera, Semilla M. wrote:
>Thanks so much, Richard. I will try this. :)
Another approach (there are many) is to use commands like
. gen touse = !missing(y, x1, x2, x3, x4)
. logit y x1 x2 if touse
That is, you create a variable, e.g. touse, which is coded 1 for the cases
you want to analyze and 0 otherwise. In this case, you are restricting
your analysis to cases with complete data on y, x1, x2, x3 and x4. You can
also make the selection more complicated if you need to, e.g. limit it to
females with nonmissing data.
The above approach, of course, assumes that you want to do listwise
deletion of missing data. That is probably the most common strategy, but
there are others, e.g. impute values for the missing data.
Finally, there is a -force- option on the -lrtest- command that might make
your original error go away, but you probably don't want to use it, i.e. it
probably won't give you the right results.
*
* 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/
<<winmail.dat>>
