Back to the main page.

Bug 896 - plotting of coherence statistics witg ft_multiplotTFR fails

Status CLOSED FIXED
Reported 2011-08-24 15:21:00 +0200
Modified 2011-08-31 14:50:30 +0200
Product: FieldTrip
Component: plotting
Version: unspecified
Hardware: PC
Operating System: Windows
Importance: P1 normal
Assigned to: Jan-Mathijs Schoffelen
URL:
Tags:
Depends on:
Blocks:
See also:

Laura Marzetti - 2011-08-24 15:21:34 +0200

see attached data file, it fails both for the sparse and the full representation.


Jan-Mathijs Schoffelen - 2011-08-24 15:24:30 +0200

please add attachment. so far I have never encountered problems. did you use connectivityanalysis?


Laura Marzetti - 2011-08-24 15:31:34 +0200

the file is too large as attachment, but can be found on mentat in ~laumar/plotting_bug.mat


Jan-Mathijs Schoffelen - 2011-08-24 16:08:45 +0200

please relax the reading permissions on your home directory, I cannot access it (being from a different group)


Laura Marzetti - 2011-08-24 16:48:34 +0200

(In reply to comment #3) > please relax the reading permissions on your home directory, I cannot access it > (being from a different group) I had some problems with that so I moved the plot_bug.mat to ~/common/temporary/laumar. I hope that will work!


Jan-Mathijs Schoffelen - 2011-08-24 16:53:20 +0200

for me stat_coh_full works -> cfg.refchannel should not be numeric, but the label of the channel (I guess in this case 'fastica004' stat_coh has strange dimensionality in the functional data, shouldn't this be reshaped to begin with? did you have a look at ft_connectivityplot? currently it does not support TF-data, but it is relatively straightforward to extend the functionality there


Jan-Mathijs Schoffelen - 2011-08-24 16:54:38 +0200

for the stat_coh variable, it's the ft_prepare_layout function which fails -> it can not deal with an absent label-field (because there's only labelcmb in the data).


Robert Oostenveld - 2011-08-25 09:17:25 +0200

Regarding reshaping: you are right, but that does not make a difference for the bugs presence. Regarding the numeric cfg.channel, that is according to the specification of FT_CHANNELSELECTION -> fix needed. Regarding sparse/full representation, that is something that the FT function should handle -> fix needed? Regarding the absence of stat_coh.label, that is something that the FT function should handle -> fix needed. The input structure, either sparse (after some reshaping) or full, is a technically correct description of a T data structure and hence it should work. If the data structure is not supported, then ft_checkdata should inform the user. Right now ft_checkdata and ft_datatype both don't complain.


Jan-Mathijs Schoffelen - 2011-08-25 09:38:12 +0200

The issue with the numeric cfg.refchannel is now fixed. Wrt to the other problem, my suspicion is that the fix is needed in ft_prepare_layout


Jan-Mathijs Schoffelen - 2011-08-25 10:01:02 +0200

I now also fixed the second problem. Please first reshape the data in the sparse representation, so that it reflects the dimord, i.e. stat_coh.stat = reshape(stat_coh.stat, [size(stat_coh.labelcmb,1) numel(stat_coh.freq) numel(stat_coh.time)]);


Robert Oostenveld - 2011-08-25 10:20:02 +0200

I created a test script and copied the data (reshaped) to home/common. See http://code.google.com/p/fieldtrip/source/detail?r=4012 and http://code.google.com/p/fieldtrip/source/detail?r=4013