Date:         Mon, 7 Aug 2006 17:49:14 -0400
Reply-To:     Ted Barker <ted.barker@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Ted Barker <ted.barker@GMAIL.COM>
Subject:      PROC MI & MIANALYZE for MEANS and LSMEANS
Content-Type: text/plain; charset=ISO-8859-1; format=flowed

Greetings,

I hope someone may have knowledge I am lacking,!

I am interested in combining MEANS STD and LSMEANS, STDERR estimates from
PROC GLM with a 3-way interaction (class variables). Is it possible to
output these estimates (PROC GLM) from multiple imputations (PROC MI) and
combine in PROC MIANALYZE?

I am trying to avoid parameter estimates via GLM.

Any help on this matter will be greatly appreaciated!

The PROC GLM syntax is something like this:


proc GLM data=impute5 ;
  class theft phys ASEXE_2;
  model FBLTOT FEVES01 FVCCS02  =  phys|theft|ASEXE_2;
  format theft steal. phys aggr. ;
  lsmeans phys|theft|ASEXE_2 / stderr pdiff tdiff ;
  means phys|theft|ASEXE_2 ;
  MANOVA h=_all_;
  by _imputation_ ;
  ods output  LSMeans=MEANSLS
            MEANS=RawMEANS;
RUN ;QUIT ;

--
Ted Barker, Ph.D
Department of Psychology
University of Montreal