[FieldTrip] statistics on non-event-related fields

[Eelke Spaak]

Dear Marco,

The FieldTrip statistics routines support permutation of condition
labels among units of observation. I guess in your data you don't
really have 'units of observation', i.e. you have continuous data of
one (or several) subjects. In that case I would recommend taking care
of the statistics outside of FieldTrip, for instance by using a
randomisation approach based on shifting time series of different
channels by different, random, amounts. The coupling values obtained
by these shifted time series can serve as a distribution under the
null hypothesis of no coupling. The usual cluster machinery can then
be applied (i.e. combining above-(nonparametric)threshold
time-frequency-channel voxels into cluster candidates, compute cluster
statistics per randomization, and compare the observed cluster
statistic to the randomization distribution). You would also need to
write this yourself, but it should not be very difficult. The
mex-files bwlabel and spm_bwlabel (distributed with FieldTrip) are
very useful; they give index labels to connected clusters in a binary
matrix.

Note, however, that there is an important caveat with the approach I
describe here. The time shifting per channel also destroys the
between-channel structure in your data that is due to electric volume
conduction. So even if you find significant connectivity by this
approach, although the connectivity would be 'real' in a sense, it
still might not be meaningful if you do not account for this volume
conduction. This is something to think about apart from the
statistics.

Hope this helps.
Best,
Eelke

On 21 June 2013 14:26, Marco Porta <marco.porta88 at gmail.com> wrote:
> Dear Users and Eelke,
> I have spontaneous LFP data recorded intracranially. I'm interested in
> studying phase correlation between sensors and assess such correlation
> within single subjects studies. Is it possible to study statistical
> significance in such correlation study or should i implement my own
> statistic?
> Thanks,
>
> Marco
>
>
>
>> Dear Marco,
>>
>> What do you mean exactly with "non event-related fields"? I presume
>> there is some structure in your data that you want to consider as the
>> independent variable of interest, right? Some more information on what
>> you want to do would help us to help you.
>>
>> Best,
>> Eelke
>>
>>
>>
>>
>> Dear Fieldtrip experts,
>> I have a question regarding the statistics. How can I statistics on non
>> event-related fields in a between-trials.
>> Thanks,
>> Marco
>
>
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip