Back to the main page.

Bug 2381 - error in ft_sourcestatistics?

Status REOPENED
Reported 2013-11-19 23:02:00 +0100
Modified 2015-05-07 13:30:57 +0200
Product: FieldTrip
Component: core
Version: unspecified
Hardware: PC
Operating System: Windows
Importance: P3 normal
Assigned to: Eelke Spaak
URL:
Tags:
Depends on:
Blocks:
See also:

Vitoria Piai - 2013-11-19 23:02:38 +0100

% I'm not sure whether this is a bug but following Jan-Mathijs' request, here's % some sample data for which I'm getting the error with example code, including % the error I'm getting. % The data can be found at common/temporary/4Bugzilla/AVG.mat % The code + error: nsubj = 15; cfgSt.method = 'montecarlo'; cfgSt.statistic = 'depsamplesT'; cfgSt.parameter = 'avg.pow'; cfgSt.correctm = 'cluster'; cfgSt.numrandomization = 500; cfgSt.alpha = 0.0500; cfgSt.clusteralpha = 0.0500; cfgSt.dim = AverageSource1.dim; cfgSt.design(1,:) = [1:nsubj 1:nsubj]; cfgSt.design(2,:) = [ones(1,nsubj) ones(1,nsubj)*2]; cfgSt.uvar = 1; cfgSt.ivar = 2; cfgSt.correcttail = 'prob'; stat = ft_sourcestatistics(cfgSt, AverageSource1, AverageSource2); % Reference to non-existent field 'avg'. % % Error in getfield (line 37) % f = s.(deblank(strField)); % deblank field name % % Error in getsubfield (line 44) % s = getfield(s, t); % % Error in statistics_wrapper>get_source_trial (line 544) % tmp = getsubfield(varargin{i}.trial(j), cfg.parameter); % % Error in statistics_wrapper (line 204) % [dat, cfg] = get_source_trial(cfg, varargin{:}); % % Error in ft_sourcestatistics (line 107) % [stat, cfg] = statistics_wrapper(cfg, varargin{:});


Jan-Mathijs Schoffelen - 2013-11-20 09:04:31 +0100

It runs through just fine if you change cfg.parameter into 'pow'. (although the resulting t-maps look funny: check this and verify that this is or is not what you would expect; if it isn't I suspect an issue related to the geometrical units, which Robert has been fooling around with of late).


Eelke Spaak - 2015-02-11 16:56:51 +0100

The test script started failing, it has to do with getdimsiz() (and hence ft_datatype_source) not properly understanding the specific input data for this bug. I'm working on it.


Jan-Mathijs Schoffelen - 2015-05-07 13:25:27 +0200

I was going through the list of failing test functions as per the dashboard, didn't notice that this one was already on someone's radar. It seems that ft_datatype_source is choking on converting the data in the file on disk into the newest source structure


Jan-Mathijs Schoffelen - 2015-05-07 13:30:57 +0200

the data on disk both seems to contain an 'avg' and a 'trial' field. If I do rmfield(source1, {'avg' 'var'}) and likewise for source2 (these are the variables in the file), the function stops failing. I suspect that the input data was obtained with a outdated version of ft_sourcegrandaverage that doesn't fit our current definition of data structures anymore. I would vote for cleaning up the test data for this bug, so that the regression function becomes functional again, rather than trying to make this backward compatible.