See also a thread about six weeks ago, especially
Richard Goldstein's posting on 16 April.
Nick
[email protected]
Trudy Ann Cameron
> You can create a dummy variable for "have_variable" that
> takes a value of 1
> if the data are non-missing and zero otherwise. Then you can
> interact this
> variable with the continuous variable. The slope that you
> estimate on the
> interaction with the continuous variable is a slope CONDITIONAL on the
> availability of data for that variable. The coefficient on the dummy
> variable is the difference in the expected value of Y
> according to whether
> the data on your continuous variable is missing or not.
>
> There is not much you can do if the continuous variable is
> your dependent
> variable, unless you know that 99999 is a top-coded variable,
> or something
> like that.
>
> The alternative is to estimate a selection correction model using the
> "have_variable" dummy as the selection variable, and to
> correct the "main"
> regression for any selection bias.
>
> Trudy
>
> At 12:17 PM 6/2/2004, you wrote:
> >Hi, I have continuous variables that are coded 99999 or
> 77777 for unknown
> >or refused to answer. How do I handle this? I don't want to
> treat them as
> >missing, if possible.
*
* 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/
