Hi,
I'd like to add a 4D voxelwise regressor to my model. I see the -vef and -ven options in film_gls. I'm not sure of their proper use, though, since my first trial didn't go as planned. This is, of course, at a subject level; I am starting with a filtered_func_data and just running stats usually at this point.
1. I have a 7 x EV glm to start with; there is 1 custom/single column EV and 6 confound EVs in a file. I also have an image based 4D EV nifti file ready to use.
2. When I ran film_gls without the -vef/-ven options, I get pe1 - pe7 in the stats directory as expected.
3. When I ran film_gls and used "8" for -ven and the 4D file for -vef. I received this error: "voxelwiseEvTarget is greater than number of design EVs". I thought that perhaps 0-based counting was the problem, so I
4. Ran film_gls with "7" for -ven, and the 4D file for -vef. The program ran and completed. However, my results were a very similar pe1-pe7, so I'm not sure what happened, or if my 4D file was used
So, how do I include the voxelwise EV properly? Do I need to enter a dummy EV at the design matrix level and then overwrite it? If so, are there restrictions on the dummy EV?
A second question, just to confirm, when I have primary EVs as well as confound EVs, do pe1-pen correspond to the n primary EVs, and the higher numbered EVs pe<n+1> - pe<n+m> correspond to the m confound EVs?
Thanks,
Mike Rohan
|