Back to the main page.
Bug 1249 - inconsistent grad-structure in output after ft_componentanalysis and ft_rejectcomponent
| Status | CLOSED FIXED |
| Reported | 2012-01-10 16:31:00 +0100 |
| Modified | 2013-02-23 20:56:21 +0100 |
| Product: | FieldTrip |
| Component: | core |
| Version: | unspecified |
| Hardware: | PC |
| Operating System: | Mac OS |
| Importance: | P3 normal |
| Assigned to: | Jan-Mathijs Schoffelen |
| URL: | |
| Tags: | |
| Depends on: | |
| Blocks: | 1338 |
| See also: |
Jan-Mathijs Schoffelen - 2012-01-10 16:31:23 +0100
Jan-Mathijs Schoffelen - 2012-01-10 16:39:16 +0100
The issue is that after backprojecting the data to sensor space the number of labels does not match the number of chanpos and chanori anymore. Moreover, all chanpos and chanori are NaN, but I believe that this was intended. I guess that the number of chanpos positions are all channels + the number of components removed To reproduce: see test_bug1249.m
Eelke Spaak - 2012-03-14 12:10:33 +0100
Would it be an idea to just keep the original grad structure after ft_componentanalysis? Obviously, it should *not* be called comp.grad, but something like comp.topograd (analogous to comp.topolabel, which is strictly speaking also not applicable to the data anymore). Let's discuss during the meeting.
Jan-Mathijs Schoffelen - 2012-09-19 14:56:35 +0200
This has been discussed quite extensively in a past meeting. The exact outcome of the discussion is lost but JM will decide what to do...
Roemer van der Meij - 2012-10-10 14:47:11 +0200
This has been discussed during the bug binge, and the exact outcome of the discussion was not entirely lost. JM will work on this, keeping in mind the recent work to ft_rejectcomponent. Suspected cause = keepunused = 'yes' at line 176 of ft_rejectcomponent. It could be that it keeps not only the reference channels, but also the 'components', 'grad-wise'.
Jan-Mathijs Schoffelen - 2012-10-12 13:55:01 +0200
The cause of this bug has been tracked to an error in ft_apply_montage, where the subselection of chanpos was omitted.