Comparison between modeled and unmodeled results - SAS proc GLM
I'm using SAS proc GLM to compare two groups (test/control) and noticed something I don't really understand.
I ran two versions of the analysis, one model controlling for age and gender and one that does not. Looking at the results side-by-side, I noticed that the means in the adjusted analysis was slightly lower than the unadjusted anlalysis across both Test and Control groups and I don't understand why.
For example, in the unadjusted analysis, the means were:
Test - 0.20
Control - 0.21
and in the adjusted analysis if was
Test - 0.19
Control - 0.20
I expected that the adjusted analysis would control for differences between the two groups but that the total across both would be the same as unadjusted. Any help would be greatly appreciated.