 |
 |
Dear Ankit, > I am relatively new to Feat and FSL and am having issues with my > design > matrix. I have gone through the tutorials and have still not been > able to > find a solution to my problem. I understand that with the increase > in the > number of EV's FSL is going to complain about rank deficiency, so > what do i > do in order to not get this warning and yet have all the EV's in my > lower > level design. I am having a total of 35 EV' s and the warning i get is > > "Design matrix is rank deficient - ration of min:max eigenvalues is > SVD of > matrix is 1.075e-17" This means that there are 1. EV's that are so similar to each other so as to be virtually indistinguishable or 2. Some set of EV's that can be combined in such a way as to be indistinguishable to some other EV > Should I just include a few EV's in my lower level design (say 5 > instead of > 35) run the analysis and then write a script to append other EV's in > the > same design ( i.e. design.fsf file) ...but i am not sure if that > would help > either...? I am not sure what you mean here. Did this occur when setting up a 2nd/ higher level analysis after you have already successfully run first level analyses with 35 EV's ? Or do you get this problem when trying to run you 1st level analysis with 35 EV's? In any case I would advice you to think through your design and see if all 35 EV's are really necessary. It is not so much the loss of df (we typically have plenty of that at the first level), but more an issue of how unique and interpretable they are. When you have strong dependencies in your design it often becomes non-obvious what you are actually testing for. Good Luck Jesper