Hello,

I am trying to do a multiple regression with 3dRegAna and I am a bit confused about the model option. Based on what I have read from the AFNI documentation, and to a certain extent the message board, I understand that the regression coefficients and their t statistics should not be affected by the model specification. However, I am finding these values are in fact different (albeit slightly) when I change the model option. Specifically, I am finding that the t statistics are more robust when using the -model 0:1 option (relative to -model 1:0 2). Below are the two options I am running:

3dRegAna \
-rows 23 \
-cols 2 \
-xydata 12.300 11.96 '/data/.../K0042-0407-WM_all-reg-CR-b6-masked-scaled-resam3+tlrc[26]' \
-xydata 5.3000 10.72 '/data/.../K0058-1007-WM_all-reg-CR-b6-masked-scaled-resam3+tlrc[26]' \
...
...
...
-xydata 19.600 15.22 '/data/.../K0529-0709-WM_all-reg-CR-b6-masked-scaled-resam3+tlrc[26]' \
-rmsmin 0 \
-model 1 : 0 \
-bucket 0 S2BK-Vig_Testosterone+Age_regression_23female_subjs


3dRegAna \
-rows 23 \
-cols 2 \
-xydata 12.300 11.96 '/data/.../K0042-0407-WM_all-reg-CR-b6-masked-scaled-resam3+tlrc[26]' \
-xydata 5.3000 10.72 '/data/.../K0058-1007-WM_all-reg-CR-b6-masked-scaled-resam3+tlrc[26]' \
...
...
...
-xydata 19.600 15.22 '/data/.../K0529-0709-WM_all-reg-CR-b6-masked-scaled-resam3+tlrc[26]' \
-rmsmin 0 \
-model 1 : 0 2 \
-bucket 0 S2BK-Vig_Testosterone+Age_regression_23female_subjs


What could be a reason for this discrepancy?

Thank you!
Gabriela