Hi Megan and Conn Experts

I am also confused by the F-test DOF for a while now. I think similar questions have been asked several times over the years but unfortunately it never got replied comprehensively because most of the related posts before also had other more pressing issue in them and drawn away the focus to things other than the F-test DOF.

I have looked into the codes, particularly the conn_process.m and conn_glm.m and found out in line 4555 in con_process.m,
there is a line of code that goes as followed: ndims=ceil(size(ROI.y,1)/5); 
the size(ROI.y,1) part gives you the number of the subjects.
So in your case, it will be 31. Then 31/5 equals 6.2 and with the ceil function ndims turns out to be 7, this will later be utilized and pass into conn_glm.m and I think it is related to the resulting DOF.

In line 159 in conn_glm.m, dof= [Ns,dofe-Ns+1];
Ns will be related to the ndims, so in you case Ns would be 7.
dofe=N1-Nx;N1 is equal to the number of subjects, and Nx is equal to the number of elements in the contrast vector. 
In your case, N1 should be 31, Nx=3, so dofe=28.
Therefore the dof=[7, 28-7+1]=[7,22] respectively.
I think that is why you get 7,22 as the DOFs in the F-test.

But even with the above, I still don't understand the rationale of ndims=ceil(size(ROI.y,1)/5) in the first place. Especially as why this would make the F-test  an omnibus test for all the possible connection for a given seed. To me, it does not seem to take into account the total number of possible connections in this equation. So I think I must have missed something.

I would really love to have the CONN experts lift me out of this confusion I am having for a while.

Thank you very much!

Chen-Chia
Originally posted by Megan Fitzhugh: