It could be caused by several things, but in this case I think the problem is with the mc-parameters as you suggest. fsfast orthogonalizes the parameters and then uses the top three, but it may be the case that the 3rd is badly conditioned, but I've never seen that. You can include the top 2 with:
-extreg mcextreg -nextreg 2
But this will force you to use the top 2 in every analysis.
Can you send me the fmc.mcdat file?
doug
Hans Op de Beeck wrote:
Hi all,
When I run selxavg-sess, I sometimes get the following error message at the beginning of the algorithm (right after I get the Residual Error Estimation):
ERROR: paradigm is ill-conditioned (1.17293e+07) Check your paradigm file for file for periodicities
My paradigm is fairly simple. It is a block-design, I use -gammafit, and there are 3 conditions that I present in 3 different orders across runs (e.g., in one run it might be 0 1 2 3 0 3 2 1 0). I always use the same paradigm-files, and this error message doesn't appear always.
In general I include the motion-correction parameters as covariates (flag -mcextreg in mkanalysis-sess.new). The problem is easily solved by not including the motion-correction parameters in the analysis. Can this error message be caused by a strong correlation between 2 of my mc-parameters ? What is the best solution in that case ? Maybe to compute a lower number of principal components whenever the problem occurs ?
Thank you for your help,
Hans P. Op de Beeck Kanwisher lab
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer