Re: st: Recoding missing variables

[David Kantor <kantor.d@att.net>]

At 11:23 PM 7/15/2006, Ashwin wrote:
> Thanks for the reply. I do not want to drop the entire
> observation as it has nonmissing values for some other
> variables. For examples,
>
>
>         var1     outcome1 var3   var4
> obs1     1         2       7     99
> obs2     99        99      3      1
> obs3      0         2       5     4
> obs4      1         0       99    99
>
>
> For analyses of var1 , i want to include obs1, 3, 4
> and for var 3 - obs 1,2, and 3.
>
> However, when i run the regression for outcome1
> against var1, var3 or var4, I do not want the "99" to
> be included as a legitimate coding for those variables
> and hence influence the odds ratio estimate.
>
> Hope that clarifies my question a bit.
> [...]
Just recode 99 to missing, as some other people have advised you to 
do. (You can use -mvdecode- or -recode- or -replace-.)

Then, in any regression, only those observations with missing values 
in any of the variables used in that regression are omitted.

Then, for example, in
 regress outcome1 var1
observation 2 will be omitted, and in
 regress outcome1 var3
observations 2 and 4 will be omitted.

This is probably what you had in mind, but this has the slight 
"problem" that as you tweak your covariate set, the observation set 
may change as well.

I hope this helps.
--David

*
*   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/