Back to the main page.
Bug 1593 - number of channels can be mismatched with neighbours in ft_statistics_montecarlo with clusters
|2012-07-05 10:34:00 +0200
|2019-08-10 11:55:41 +0200
Robert Oostenveld - 2012-07-05 10:34:04 +0200On 29 Jun 2012, at 4:56, Arnaud Delorme wrote: - there is no check that the "chan" dimension contains the correct number of channels for multiple comparisons (in other words that it is the same size as the neighbouring matrix). I have run several simulations where Fieldtrip detected that (I had my channels defined in the "freq" dimension and Fieldtrip performed the cluster correction there - which is strange because for freqs, Fieldtrip should perform a classical correction not based on channel neighbours - maybe it did?). Anyway, it would be good if cluster correction for channels could only be done in the channel dimension and if there was a check that the neighbouring matrix has the same size as the chan dimension. I can investigate more to try to get you a reproducible example if you want.
Robert Oostenveld - 2012-07-05 10:36:03 +0200Hi Arno, Can you provide a test script that demonstrates the problem? It should be possible to have channels that are not neighboured, whereas other channels are (e.g. including a bipolar HEOG in the statistics, whereas all commonly referenced EEG channels have neighbours) I can imagine it being needed to check that the neighbourhood definition specifies the neighbours (or the absence thereof) for all channels that are included in the test
Robert Oostenveld - 2012-07-05 10:43:15 +0200Hi Arno, I have assigned this one to you for the moment. Please assign it back to the joint development user once you are done with it. Robert
Arnaud Delorme - 2012-07-05 21:16:41 +0200I did more tests and it did generate an error ??? Error using ==> findcluster at 59 invalid dimension of spatdimneighbstructmat I must have been confused. Arno
Robert Oostenveld - 2019-08-10 11:51:43 +0200This closes a whole series of bugs that have been resolved (either FIXED/WONTFIX/INVALID) for quite some time. If you disagree, please file a new issue describing the issue on https://github.com/fieldtrip/fieldtrip/issues.