I will appreciate any input you may give.
I am runing glm procedure, but I get one more test significant in contrast
than in pairwise multiple comparison. I don't know why. Below is the code
and output.
proc glm data=out.final;
class dx;
format dx dx4x.;
model &v =dx;
lsmeans dx;
means dx/tukey;
contrast "demented vs normal" dx 1 0 -1;
contrast "demented vs mci" dx 1 -1 0;
contrast "mci vs normal" dx 0 1 -1;
run;
1. Least Squares Means;
Least Squares Means
Q1_PARTAPP
dx LSMEAN
DEMENTED 6.00000000
MCI 5.71428571
NORMAL 4.86486486
2. Contrast
Dependent Variable: Q1_PARTAPP CONSIDER TRIAL OF APP MED
Contrast DF Contrast SS Mean Square F Value
Pr > F
demented vs normal 1 15.60294840 15.60294840 7.09
0.0094
demented vs mci 1 0.89440994 0.89440994 0.41
0.5256
mci vs normal 1 11.49985150 11.49985150 5.23
0.0249
3. Pairwise Multiple comparison
Difference
dx Between Simultaneous 95%
Comparison Means Confidence Limits
DEMENTED - MCI 0.2857 -0.7845 1.3559
DEMENTED - NORMAL 1.1351 0.1171 2.1532 ***
MCI - DEMENTED -0.2857 -1.3559 0.7845
MCI - NORMAL 0.8494 -0.0379 1.7368
NORMAL - DEMENTED -1.1351 -2.1532 -0.1171 ***
NORMAL - MCI -0.8494 -1.7368 0.0379
