Dear experts,
I've been reading different suggestions proposed to bring a FIR analysis to group level, considering the difficulty to carry F tests to higher level, etc.
I have been thinking about another way to do it (maybe it was already mentioned and I skipped it! excuse me if I did), and I'd like to know your opinion about it.
Let's suppose I have run a FIR analysis with 8 time bins in 17 subjects. Using FEAT, I run a one sample t test of the 17 subjects. I end up with a gfeat directory that contains 8 feat directories (cope1, cope 2... cope 8, one for each FIR), each of them with a filtered_func_data file, which is what you usually put in randomise as input.
My question is: for an overall effect, couldn't I just merge all filtered_func_data files in one file (giving a 4D file with 17x8=136 "time" points) and run randomise with -1 option? According to my assumption, maybe wrong, I'll get those voxels with a COPE value different than zero considering all time bins.
In the same line, if I want to compare my 17 subjects in condition A with my 17 subjects in condition B, I could just merge the 17x8x2=272 filtered_func_data files and run randomise with design.mat and design.con files taken from a paired t test analyses (being extremely careful with the order of the files).
Does it make any sense? Thanks a lot!
Javier
|