clusterstats empty cfg.neighbours, cfg.neighbours example
arno at CERCO.UPS-TLSE.FR
Mon Mar 8 18:05:33 CET 2010
I was running some correction for multiple comparisons using the
cluster method you have developed and I set the number of neighbors
for each channel to 4.5 when constructing the channel neighbor matrix
(which seem meaningful to me as each channel has in general 4
neighbors in the 4 directions). I did work and returned relevant
results. I am now thinking of making a link in EEGLAB to use that
function but I want to be sure that my arbitrary choice in the number
of channel neighbor is a good one.
Thanks a lot,
On Mar 8, 2010, at 8:58 AM, Robert Oostenveld wrote:
> Hi Chetan,
> Sorry about that, the bookeeping of the data dimensions is a bit
> messy, as reflected in the cfg.dim which is pre and post appended all
> the time.
> On 4 Mar 2010, at 16:51, Chetan Sharma wrote:
>> I am still running into bugs when doing the freq_time clustering. I
>> am testing for coherence between the hemispheres during a task, and
>> now it runs into problems at clusterstat() at line 175.
> what is your statfun?
> Channel clustering is not supported for all possible statistics, and
> especially has limits in the case of bivariate measures such as
>> The command is:
>> posclusobs = findcluster(reshape(postailobs,
>> The problem is that postailobs has size [nfreq*ntime, 1], while the
>> cfg.dim is [nchan nfreq ntime], where nchan=2 in this case. It seems
>> like it wants to reshape it into size [nfreq, ntime, 1]. Is this
>> indeed the case?
>> And could someone help me out by explaining a bit more about when
>> one would choose for cfg.clusterthreshold the different values,
>> 'nonparametric_common', 'nonparametric_individual', and 'parametric'
>> when testing for coherence in different brain areas? I have read
>> over the documentation and the code but am still somewhat unsure.
> parametric uses the known parametric distribution of the statistic,
> this has to be supported by the statfun.
> nonparametric_common estimates a threshold from the randomization
> distribution. The threshold is common to all channe-time-frequency
> nonparametric_individual estimates an individual threshold from the
> randomization distribution for each channe-time-frequency point.
> best regards,
> The aim of this list is to facilitate the discussion between users
> of the FieldTrip toolbox, to share experiences and to discuss new
> ideas for MEG and EEG analysis. See also http://listserv.surfnet.nl/archives/fieldtrip.html
> and http://www.ru.nl/neuroimaging/fieldtrip.
The aim of this list is to facilitate the discussion between users of the FieldTrip toolbox, to share experiences and to discuss new ideas for MEG and EEG analysis. See also http://listserv.surfnet.nl/archives/fieldtrip.html and http://www.ru.nl/neuroimaging/fieldtrip.
More information about the fieldtrip