help
484 Subscribers 
help > Group effects in 2nd level analysis
Showing 1-2 of 2 posts
May 25, 2018 06:05 PM | David Cochran
Group effects in 2nd level analysis
If I am comparing group differences in functional connectivity,
what is the difference between:
a) creating a "Group" variable that is 0 for controls and 1 for patients, and doing contrast [AllSubjects Group] = [0 1]
b) creating a "Control" variable that is 1 for controls and 0 for patients, and "Patient" variable that is 1 for patients and 0 for controls, and doing contrast [Patient Control] = [1 -1]
When I do the first, I have significant results at p-uncorrected 0.001 voxel-wise and p-FDR=0.05, but for the second I get no significant results.
a) creating a "Group" variable that is 0 for controls and 1 for patients, and doing contrast [AllSubjects Group] = [0 1]
b) creating a "Control" variable that is 1 for controls and 0 for patients, and "Patient" variable that is 1 for patients and 0 for controls, and doing contrast [Patient Control] = [1 -1]
When I do the first, I have significant results at p-uncorrected 0.001 voxel-wise and p-FDR=0.05, but for the second I get no significant results.
May 25, 2018 09:05 PM | Alfonso Nieto-Castanon - Boston University
RE: Group effects in 2nd level analysis
Hi David,
There is no difference at all between the two approaches (assuming that your subjects are either controls or patients, i.e. assuming that there are no additional subjects in the AllSubjects group). Both should result in exactly the same results (including same effect-sizes, same T- statistics, dof's, p-values, etc.)
I would suggest to double-check your second-level covariate definitions just to make sure that they look fine (if in doubt just send me your conn_*.mat file and I will take a quick look)
Best
Alfonso
Originally posted by David Cochran:
There is no difference at all between the two approaches (assuming that your subjects are either controls or patients, i.e. assuming that there are no additional subjects in the AllSubjects group). Both should result in exactly the same results (including same effect-sizes, same T- statistics, dof's, p-values, etc.)
I would suggest to double-check your second-level covariate definitions just to make sure that they look fine (if in doubt just send me your conn_*.mat file and I will take a quick look)
Best
Alfonso
Originally posted by David Cochran:
If I am comparing group differences in
functional connectivity, what is the difference between:
a) creating a "Group" variable that is 0 for controls and 1 for patients, and doing contrast [AllSubjects Group] = [0 1]
b) creating a "Control" variable that is 1 for controls and 0 for patients, and "Patient" variable that is 1 for patients and 0 for controls, and doing contrast [Patient Control] = [1 -1]
When I do the first, I have significant results at p-uncorrected 0.001 voxel-wise and p-FDR=0.05, but for the second I get no significant results.
a) creating a "Group" variable that is 0 for controls and 1 for patients, and doing contrast [AllSubjects Group] = [0 1]
b) creating a "Control" variable that is 1 for controls and 0 for patients, and "Patient" variable that is 1 for patients and 0 for controls, and doing contrast [Patient Control] = [1 -1]
When I do the first, I have significant results at p-uncorrected 0.001 voxel-wise and p-FDR=0.05, but for the second I get no significant results.