[Homer-users] GLM: request for documentation & possible bug

[Igor Fridman]

Hi all,

The GLM function (hmrDeconvHRF_DriftSS.m) currently lacks documentation 
aside from what is in the file's comments and in the instructional 
video.  I know folks are working hard on this, and I'd like to ask these 
questions:

- is there a particular order in which it must be called in the 
procStreamGUI dialog, i.e. after the enCrossCorrelation.  Should I 
delete the blockAvg function?

-is there any way to get the de-convolved data for the entire sample and 
before block averaging?

-is there any way to plot the HRF before & after the GLM deconvolution 
side-by-side, aside from exporting the plots/data manually

-I am getting this error: ??? Output argument "yR" (and maybe others) 
not assigned during call to "[...] hmrDeconvHRF_DriftSS". It appears to 
be dependent in the function parameters selected in EasyNIRS_ProcessOpt, 
but I've not yet identified which are to blame.

- I also get warnings like: Design matrix is poorly scaled...(RCond = 
1.448025e-015)  What does this refer to?

- There appears to be a bug in the GLM functions and many of the other 
short-separation (SS) deconvolution functions.  When searching for the 
SS channel(s), all channels with the first wavelength must be 
consecutive in the dataset, otherwise you get an out-of-bounds error 
when calling posM(lstSS,;) (i.e. in line 117 of 
hmrDeconvHRF_DriftSS.m).  What i mean is, you get an error if 
SD.MeasList = [[1,1,1,1], [1,1,1,2],[1,1,1,3],...], but not if you 
reorder the channels as SD.MeasList = [[1,1,1,1], [1,2,1,1],[1,3,1,1],...].

- I've written a small Python wrapper to convert arbitrary-sized CSV 
(excel text) files to .nirs format.  This may be useful to those who can 
export to CSV from their fNIRS machine but don't have access to matlab.  
Please message me if you need it, and I could also contribute it to the 
homer website.

Thank you

-- 
*Igor Fridman, Ph.D.
Research Scientist
Quantum Applied Science and Research, Inc.
Tel: 858.412.1851
www.quasarusa.com <http://www.quasarusa.com>*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/homer-users/attachments/20140513/8a277dee/attachment.html