[Marsbar-users] Very slow estimation

[Jiansong Xu U. <jia...@ya...>]

Dear Friends:

I'm trying calculate % signal changes of ROIS using Marsbar. It usually
works very well. But, now the estimation is very slow and takes several
hours for one ROI.  It is much slower than the estimation of SPM2. SPM2
takes about 10 minutes to estimate for the whole brain analysis.

Another issue is that for the same subject, marsbar perform estimation
separately for each ROI. But, my previous experience is that Marsbar
estimate once and than spit out % signal changes for different ROIs.

Relative to my previous design, my current design use 6 motion parameters as
covariates. I also used RobustWLS
(http://www.bangor.ac.uk/~pss412/imaging/robustWLS.html) in my design to
reduce the effects of motion.

Marsbar print following message:

Summarizing data with summary function: mean
Re-estimating covariance
Re-estimating whitening filter
Initialising parameters
Covariance estimate
Model
Temporal non-sphericity (over voxels)
......
......



Any suggestions will be highly appreciated.


Best

-- 
Jiansong Xu, M.D., Ph.D.
Associate Research Scientist
Dept. of Psychiatry
Yale Medical College
34 Park St., Room S-100
New Haven, CT 06519