Hello Experts,
I am trying to run a 3rd level feat analysis with 2 main EV's and 1 additional, voxel dependent EV.
I am running into problems when I add the voxel dependent EV into the analysis (it works fine without it).
Here is how the model is set up. There are 30 inputs and each one has been set to group 1.
The first 15 inputs have been set to 1 for EV1 and 0 for EV2, the opposite is set for the last 15 inputs.
The idea is to use the voxel dependent EV to model the between subject variance, rather than assigning the inputs to
different discrete groups.
In particular we're using either breathhold or resting state data to generate our voxel dependent EV. For each voxel we have calculated
the mean subtracted activity (activity in a voxel for a given subject - average activity in that voxel for all subjects).
However, we are missing data for a few subjects. We have attempted to model the data for the missing subjects by taking the mean of the mean subtracted activity for each voxel and creating a 'dummy' subject which we have put in place of the missing subjects.
Some of the values in the matrix are 0 and we have been getting the following warning messages:
WARNING: The passed in varcope file, tmpvarcope0099, contains voxels inside the mask with zero (or negative) values. These voxels will be excluded from the analysis.
WARNING: FLAME Stage 2 has given an abnormally large difference to stage 1
Each of these warnings is repeated several times in the log file.
I tried to add .1 to every voxel to get rid of the zeros, but the analyses would completely crash every time.
My questions are the following: How concerned should I be about these warning messages? Are they indications that something is wrong with the analysis or how I have set it up? If they are a significant concern, how can I fix them?
Thanks for any answers!
|