--- Stijn Ruiter <[email protected]> wrote:
> * In the multiple imputation literature, it is suggested that the
> imputation procedure should allow binary covariates to be imputed
> with values ranging between 0 and 1. This means that missings on e.g.
> gender might indeed be imputed as 0.79 or 0.13. Apparently, we should
> think of these people as half-man, half-woman.
I have seen it recomended as a `quick fix', but not as a proper
recomondation in the MI literature I know. For good reason, with
multiple imputation you think that given the observed variables you can
estimate which values for the missing values are more likely than
others. This is information is summarized by a probability
distribution, the posterior predictive distribution. The imputations
are generated by randomly drawing from this distribution. Most
reasonable models would assign the value half-man, half-women a
posterior (and prior) probability of zero, and it would thus be
imposible for such an imputation model to generate imputed values of
any other values than either man or women.
Occationally I have seen draws from a (truncated) normal distribution
for a dichotomous variable seen promoted as a reasonable approximation
because no software is said to be available to do a more proper job.
However, that is no longer a good excuse. A lot of effort has been
spent on developing sofware that will produce discrete imputations for
discrete variables, for instance -ice- in Stata, mice in Splus and R,
IVEware in SAS, cat and mix in Splus and R.
Hope this helps,
Maarten
-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands
visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434
+31 20 5986715
http://home.fsw.vu.nl/m.buis/
-----------------------------------------
Send instant messages to your online friends http://uk.messenger.yahoo.com
*
* 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/
