st: incomplete variables in the RHS of the imputation model

[Gonzalo Rivero <griverorz@gmail.com>]

Hi everyone,

I am doing multiple imputation on survey data using mi impute mvn.
However, the variables I want to include in the RHS of the imputation
model have not been completely observed themselves. Therefore, adding
these variables to the model would create more plausible imputations,
but they would also reduce the number of filled cases.

I was wondering whether there is any literature on this subject or
standard advice for this kind of situation. In particular, does it
make any difference if add them to the LHS?

Many thanks,

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