Back to the main page.

Bug 1460 - clusterbasedpermutation freq plots nans (i.e. nothing)

Reported 2012-05-08 13:14:00 +0200
Modified 2012-08-23 14:02:04 +0200
Product: FieldTrip
Component: core
Version: unspecified
Hardware: PC
Operating System: Windows
Importance: P3 normal
Assigned to: Jörn M. Horschig
Depends on:
See also:

Jörn M. Horschig - 2012-05-08 13:14:02 +0200

I just tested the cluster based permutation tutorial for time-frequency data. This is about the actvsBlt plot I see nans in the clusterplot (i.e. completely grey), which is caused by a channel being completely 0 (which causes a std of 0 in statfun, and division by 0 = no good). Why does that suddenly constitute a problem? Seemed to have worked fine... or has the channel not been 0 before, or has been removed? (actually, the problem is that griddata cannot interpolate nans and then sets eberything to 0, but I assume that that function did not change, because it's Matlab built-in)

Lilla Magyari - 2012-05-08 13:38:09 +0200

(In reply to comment #0) Have you used the tutorial data? because in the preprocessing part some channels are supposed to be disabled: % preprocess the data = {'MEG', '-MLP31', '-MLO12'}; % read all MEG channels except MLP31 and MLO12 cfg.demean = 'yes'; % do baseline correction with the complete trial dataFIC = ft_preprocessing(cfg); I do not know if this helps.

Jörn M. Horschig - 2012-05-08 15:09:58 +0200

(In reply to comment #1) yep, but it was indeed caused by these two channels there was a bug in ft_megplanar. The function tried to interpolate on all channels available in data.grad, which still included the two dismissed channels. I fixed this now in ft_megplanar, and for the FT workshop I excluded these two channels explicitly prior to calling freqstatistics (again)

Jörn M. Horschig - 2012-08-23 14:02:04 +0200

bug closing time (