11-23-2011 02:54 PM
I'm running a multilevel model with proc mixed. I'm trying to estimate a three way interaction between homophobia (called zhomodiff), rater gender (called gender), and stimulus gender in predicting a stimuli's attractiveness (while controlling for the other things in the model: e.g. group, SCAT, PCAT).
I found a three way interaction between homophobia, rater gender, and stimulus. I was able to graph it by obtaining estimates through the LSMEANS statements. However, since the level of homophobia is a continuous variable, I am unsure if I will be able to contrast the differences in slopes based on the level of homophobia.
Here's my code thus far:
PROC MIXED covtest data=data;
CLASS id Stimulus_Number group gender sgend SCat PCat age race relationship_status;
MODEL Attractiveness = zhomodiff | gender | sgend group SCat | PCat | Sgend /solution DDFM=SATTERTH;
REPEATED Stimulus_Number /SUBJECT=id TYPE=cs;
LSMEANS Gender*SGEND/ AT zhomodiff = 1 pdiff;
LSMEANS Gender*SGEND/ AT zhomodiff= 0 pdiff;
LSMEANS Gender*SGEND/ AT zhomodiff= -1 pdiff;
Is there anyway to contrast the significance between two slopes? Is there any way to examine the statistical difference between two slopes?