convergence error svd
alenarto
Posted on 06/02/08 20:15:32
Number of posts: 41
Any thoughts? This is a new error for me. It occurs irrespective of if I try a bhvr or task PLS (I thought maybe it was bhvr data producing null values). I am looking at the raw datamats but this takes a while. One oddity in my dataset set up - 1/2 of the conditions are normalized to the first time point, and the other 1/2 are normalized to -3 time point - it shouldn't matter but I mention it!
Agatha
Loading datamat: 18 out of 18
??? Error using ==> svd
NaN or Inf prevents convergence.
Error in ==> fmri_perm_behav at 60
[brainlv,s,behavlv] = svd(stacked_datamatcorrs',0);
Error in ==> fmri_pls_analysis at 418
[brainlv,s,behavlv,brainscores,behavscores,lvcorrs, ...
Error in ==> fmri_analysis_ui>ExecutePLS at 1787
fmri_pls_analysis(PLSoptions.profiles, 'BEHAV', ...
Error in ==> fmri_analysis_ui at 90
ExecutePLS;
??? Error using ==> fmri_analysis_ui('RUN_BUTTON_PRESSED');
Error using ==> svd
NaN or Inf prevents convergence.
??? Error using ==> waitfor
Error while evaluating uicontrol Callback
Untitled Post
I'm Online
jshen
Posted on 06/03/08 16:16:48
Number of posts: 291
If you are running Behavior PLS, make sure you have at least 3 subjects, and also make sure the behavior data are not all the same. Otherwise, I can only imagine that the brain data you got is very different. Please check the brain data also.
Untitled Post
alenarto
Posted on 06/05/08 17:54:55
Number of posts: 41
Thanks Jimmy! The brain data were incorrect created. Note I had set the reference scan DURATION (rather than onset) to -3 - producing a whole lot of NaN values in the result. The program noted that it's dividing by zero in datamat creation but it didn't choke.
All fixed.
Agatha