Date:         Wed, 19 Jul 2000 17:15:21 -0700
Reply-To:     Cassell.David@EPAMAIL.EPA.GOV
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "David L. Cassell" <Cassell.David@EPAMAIL.EPA.GOV>
Subject:      Re: Multiple Imputation Macro
Content-type: text/plain; charset=us-ascii

There's no need for a macro if you can wait a bit.  PROC MI and
PROC MIANALYZE are supposed to be experimental in version 8.1 .
They're based on Rubin's 1987 text "Multiple Imputation for Nonresponse
in Surveys".  The idea is that you use PROC MI to generate M datasets
of randomly imputed values from your original dataset; then you run your
intended analysis on each; then you use PROC MIANALYZE to combine
these results to incorporate the effects of unknown data.

Note that you don't get a single set of data with the holes filled in using
'spurious' data.  You get the final analysis, with variances adjusted for
the effect of all that missing data.  If that is not what you want, you may be
asking the wrong question.  After all, imputed data can have significant
impacts on your results, and may not be the right thing to do...

HTH,
David
--
David Cassell, OAO Corp.                           Cassell.David@epa.gov
Senior computing specialist
mathematical statistician