Date:         Mon, 12 Mar 2001 13:19:48 +0300
Reply-To:     Vasya Dovbnya <Dovbnya@ASGL-RL.SPB.RU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Vasya Dovbnya <Dovbnya@ASGL-RL.SPB.RU>
Subject:      'Where' statement vs 'Contrast' statement
Comments: To: SAS-L@LISTSERV.VT.EDU

I have a problem in choosing correct way to perform
repeated measures ANOVA.
Should I use 'contrast' statement or 'where' statement for
comparisons of the Group*Time effect with a control group?

Here is my example.
I took it from examples for PROC GLM
(Example 30.7: Repeated Measures Analysis of Variance) and
modified a little bit.
It consists of 2 parts. Part 1 uses 'contrast' statement,
Part 2 uses 'where' statement.
________
Part 1

proc glm;
   class Depleted;
   model Histamine1 Histamine3 Histamine5 = Depleted / nouni;
   contrast "1-2" Depleted 1 -1 0;
   contrast "1-3" Depleted 1 0 -1;
   repeated Time 3(1 3 5) polynomial / summary printe;
quit;

Output for this procedure shows the following:

Source DF Type III SS Mean Square F Value Pr > F G - G H - F

Time*1-2 2 0.17222937 0.08611468 0.46 0.6391 0.5330 0.5685
Time*1-3 2 1.60621944 0.80310972 4.25 0.0262 0.0554 0.0451

________
Part 2

proc glm;
where Depleted="N" or Depleted="Y";
class Depleted;
model Histamine1 Histamine3 Histamine5 = Depleted / nouni;
repeated Time 3(1 3 5) polynomial / summary printe;
quit;

proc glm;
where Depleted="N" or Depleted="han";
class Depleted;
model Histamine1 Histamine3 Histamine5 = Depleted / nouni;
repeated Time 3(1 3 5) polynomial / summary printe;
quit;

Output for these procedures shows the following:

Source DF Type III SS Mean Square F Value Pr > F G - G H - F

Time*Depleted 2 0.17222937 0.08611468 1.95 0.1929 0.2215 0.2148
Time*Depleted 2 1.60621944 0.80310972 3.55 0.0481 0.0827 0.0745
________
I have no idea, why Type III SS is the same for both parts but
the F-values differ.
Please, help!

Thanks!

Vasya.