Hello,
I am running some MELODICs from the command line using data that has already been preprocessed and registered to the standard. I am using an iterative approach to test some parameters in my analysis, which entails using files that are increasing large in size (timepoints), but from the same dataset.
I encountered a problem a few times, but ubiquitously, here is an example. While trying to complete a MELODIC at a dimensionality of 70 using X number of timepoints.
Here is the error.
*****
An exception has been thrown
Logic error:- detected by Newmat: index error: requested index = 0
MatrixType = Rect # Rows = 1; # Cols = 3
*****
But all lower number of timepoints completed with no problem and a few MELODICs with an even higher number of timepoints also completed with no problem (that is the ubiquitous part).
Here is my script
*****
cd /cxfs/project2/macdonal/ICID/ICID_WMsz/ICIDretest/melodics/TRT_n27t1_dimtest/WMn27t1_d70/orderMelodics_Collected_d70/
module load fsl/4.1.8
melodic -i d70_15r_melodicIC.nii.gz -o melodic_15r_d70.ica -v --nobet --bgthreshold=3 -d 70 --vn --report --guireport=../../report.html --bgimage=bg_image --tr=2.0 --mmthresh=0.5 --Ostats
*****
Here is the last bit of code recorded in the log.txt before it fails
*****
IC map 66 ...
calculating mixture-model fit
saving probability map: metamelodic_15r_d70_4.ica/stats/probmap_66
re-scaling spatial maps ...
thresholding ...
alternative hypothesis test at p > 0.5
saving thresholded Z-stats image: metamelodic_15r_d70_4.ica/stats/thresh_zstat66
creating report page ... done
IC map 67 ...
calculating mixture-model fit
try Gaussian Mixture Model
Although I doubted more computer resources would solve the problem, I tested this anyway (64 gb -> 128gb), indeed it did not make a difference.
I have run many MELODICs with this configuration and others with different configurations and this is the first time I've encountered the problem, even in the presence of " try Gaussian Mixture Model" in the log.txt.
I am hoping you can tell me what the problem is.
Many thanks,
Krista
|