Back to PLS Help???error using SVD
Posted on 04/18/07 17:58:26
Number of posts: 0
Hi again,
I'm trying also to do a PET analysis using ANALYZE files of my MEG SAM images.
I have been able to run the results with a small number of files (3) and it worked ok. However, when I tried running it with my full sample size (n=15), the PLS analysis gives me this error:
I'm trying also to do a PET analysis using ANALYZE files of my MEG SAM images.
I have been able to run the results with a small number of files (3) and it worked ok. However, when I tried running it with my full sample size (n=15), the PLS analysis gives me this error:
??? Error using ==> svd
NaN or Inf prevents convergence.
Error in ==> pet_analysis_perm at 136
[brainlv,s,behavlv] = svd(stacked_datamatcorrs',0);
Error in ==> <a href="error:C:\MATLAB7\toolbox\pls\pet_analysis.m,331,1">pet_analysis at 331</a>
[brainlv,s,designlv,brainscores,designscores,lvcorrs, ...
Error in ==> pet_analysis_ui>ExecutePLS at 1906
[resultFile, elapsed_time] = pet_analysis(PLSoptions.isbehav, ...
Error in ==> pet_analysis_ui at 114
ExecutePLS;
Any idea why this error is occuring. Like I said, I am running the exact same analysis as I did with only a few subjects, so nothing has really changed in the analysis pipeline. Any ideas?NaN or Inf prevents convergence.
Error in ==> pet_analysis_perm at 136
[brainlv,s,behavlv] = svd(stacked_datamatcorrs',0);
Error in ==> <a href="error:C:\MATLAB7\toolbox\pls\pet_analysis.m,331,1">pet_analysis at 331</a>
[brainlv,s,designlv,brainscores,designscores,lvcorrs, ...
Error in ==> pet_analysis_ui>ExecutePLS at 1906
[resultFile, elapsed_time] = pet_analysis(PLSoptions.isbehav, ...
Error in ==> pet_analysis_ui at 114
ExecutePLS;
Replies:
Untitled Post
I'm Online
Posted on 04/19/07 15:58:54
Number of posts: 229
Steve...
During the resampling process in a behaviour analysis, it is possible to have zero variance in a behaviour measure. See Rayhan's post on 2/13/07 for a more in-depth discussion.
We have added some checks for this recently - can you double check that you have the most current version?
thanks
nancy
During the resampling process in a behaviour analysis, it is possible to have zero variance in a behaviour measure. See Rayhan's post on 2/13/07 for a more in-depth discussion.
We have added some checks for this recently - can you double check that you have the most current version?
thanks
nancy
Untitled Post
Posted on 04/20/07 16:36:19
Number of posts: 0
Hi Nancy,
I don't have any behavioral measures yet. I get the error even when I'm doing the mean-centred PLS using only my SAM (PET) images.
If the image mean is fairly close to 0, then could this produce a fairly variance free data set, once it has been normalized?
Thanks again.
S-
I don't have any behavioral measures yet. I get the error even when I'm doing the mean-centred PLS using only my SAM (PET) images.
If the image mean is fairly close to 0, then could this produce a fairly variance free data set, once it has been normalized?
Thanks again.
S-
Untitled Post
I'm Online
Posted on 04/21/07 21:32:12
Number of posts: 229
Hi Steve..
you may be right..
for this type of analysis, you have already done all the "preprocessing" necessary,
so do not normalize by anything. - the GUI will ask you if you are sure about that choice, and you are...
If you have a "brain mask" use that, otherwise, treat all voxels as brain, the PET module should remove the zeros before running the PLS analysis
Nancy
you may be right..
for this type of analysis, you have already done all the "preprocessing" necessary,
so do not normalize by anything. - the GUI will ask you if you are sure about that choice, and you are...
If you have a "brain mask" use that, otherwise, treat all voxels as brain, the PET module should remove the zeros before running the PLS analysis
Nancy
Untitled Post
I'm Online
Posted on 04/30/07 17:03:52
Number of posts: 291
If you upload all the *PETdatamat.mat files onto your ftp
server, I could look into the NaN error for you. Please let me know the ftp link.
Login to reply to this topic.
