Re: st: Re:using a flag variable for missing values in a regression

[Maarten Buis <maartenlbuis@gmail.com>]

On Wed, Mar 27, 2013 at 2:28 PM, Frank D Lopresti wrote:
> I a working with a student who's prof said in a memo "Note: when
> creating new variables, missing values should be coded as some
> arbitrary numerical value (e.g., 0) so that cases aren’t dropped in
> the regression . You can also generate a missing flag for each
> variable, coded 1 for missing and 0 otherwise, and then include the
> flags in tandem with the variables for unbiased estimates."  Is this a
> valid method for dealing with missing data  I've missed?

In general, this is not a valid way of dealing with missing values.
Here is an explanation of what happens when you use this method:
<http://www.stata.com/statalist/archive/2006-09/msg00117.html>

However, there is an exception where this method can be useful, which
is explained here:
<http://www.stata.com/statalist/archive/2011-07/msg00456.html>

Hope this helps,
Maarten

---------------------------------
Maarten L. Buis
WZB
Reichpietschufer 50
10785 Berlin
Germany

http://www.maartenbuis.nl
---------------------------------

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/