Date:         Sun, 6 Oct 1996 23:07:53 -0400
Reply-To:     Ted Floyd <floyd@LIFE.BIO.SUNYSB.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Ted Floyd <floyd@LIFE.BIO.SUNYSB.EDU>
Subject:      HELP with TEST statment in proc REG or GLM

Hi.  I'm trying to use the TEST statement to test hypotheses
about parameter estimates for interaction terms, using either
REG (ideally) or GLM (if necessary).  Consider this regression
equation:
 
    model Y = X1 X2 X3 X1*X2 X1*X3 X2*X3;
 
Now, suppose I wish to test the hypothesis that the parameter
estimates for the interaction terms are all equal to zero (i.e.,
Ho:B12=B13=B23=0).
 
How do I go about doing this?  Here are the problems I'm up
against:
 
     (a) in REG, where I had some success with the TEST
         statement for hypotheses about main effects, I
         cannot get SAS to model the interaction terms,
         period;
 
     (b) in GLM, which readily models the interaction terms,
         I reall don't understand the H= and E= options that
         appear to be required for the TEST statement.
 
I'll be really grateful for any assistance with this.
 
Thanks!
 
Ted Floyd
floyd@life.bio.sunysb.edu
 
Stony Brook, NY, USA