Hi,
I'm trying to run a regression analysis using 3dRegAna and am getting an unusual finding. I have three predictor variables: 1) age, 2) gender, and 3) age of first use of marijuana. I want to control for age and gender and look at the effect of age of first use. (The continuous variables have been centered; I realize this is not necessary here). My model is set up as such:
3dRegAna \
-rows 50 \
-cols 3 \
-workmem 1000 \
-xydata -2.2 2 -4.5 /scratch/Diffs/diff_01.nii \
-xydata 2.1 1 3.8 /scratch/Diffs/diff_02.nii \
-xydata -1.1 2 -1.8 /scratch/Diffs/diff_03.nii \
-xydata 2.9 1 -0.4 /scratch/Diffs/diff_04.nii \
-xydata 2.7 1 5.0 /scratch/Diffs/diff_05.nii \
...
-xydata 3.0 1 6.3 /scratch/Diffs/diff_50.nii \
-model 3 : 0 1 2 \
-bucket 0 Users_corr_ageuse
I do find some potentially interesting effects. I'm looking at the F-test results. The odd part is what happens when I rerun another analysis with a different predictor variable -- duration of use of marijuana. When I enter this vector into my model and rerun the analyses, I get exactly the same result as with age of first use. I tried looking for simple scripting errors but can't seem to find any. FYI, the correlation between age of first use and duration of use is small (r=-.22). But, the correlation between age and duration of use is high (.89).
Is there anything else that might be causing the results to be exactly the same?
Thanks for your help.
Jatin