[FieldTrip] Help with plotting all the sensors belonging to the significant clusters in only one scalp topography

Stephen Whitmarsh stephen.whitmarsh at gmail.com
Wed Mar 25 09:31:56 CET 2020


Dear HeYuxin,

Welcome to FieldTrip!

I saw that Julian just gave a better solution than what I was writing you,
so let me just leave you my preamble:

I assume you have done some cluster analyses over channels, frequencies and
time point. In other words, you might have clusters spanning over three
dimensions. Some channels might not be part of the cluster at some time
points, some frequencies not at certain channels, etc. etc. So you will
first have to decide what you want to look at. For instance: channels that
at *any *point and at *any *frequency were part of a cluster? That might
not be easy to interpret - what does a 'significant channel' mean in this
case? Since you say you are new in FieldTrip, I'd like to offer some
advise that might be useful here, perhaps also for others:

- Don't *start *with cluster-analyses, especially not over multiple
dimensions.
- First explore your hypothesis more directly. Your hypothesis probably is
not 'any effect at any time in any frequency in any channel'. Try to be
more precise, e.g.: at what frequencies do you expect your effect. And at
what time-period?
- If you can average over a dimension, do it. I.e. if you expect a certain
frequency band (e.g. alpha 8-14Hz) to show an effect, average over that
dimension. You can even do this in ft_freqanalysis.
- Try to avoid more than 2 dimensions in cluster-analyses. It makes
interpretation (and plotting!) much easier.
- If you absolutely need to explore 3 dimensions, you can plot the results
using ft_clusterplot.
- Reading this
<http://www.fieldtriptoolbox.org/faq/how_not_to_interpret_results_from_a_cluster-based_permutation_test/>
might help understand also how you would need to report cluster-analyses
statistics, and the limitations of their interpretation.
- To summarize: *Cluster analyses is meant to deal with the multiple
comparison problem, not a lack of precision in your hypotheses*. I do not
say this to be pedantic, but to help you avoid becoming trapped in a
circular multidimensional analyses chasing false positives, let alone
risking unconsciously p-hacking and HARKing.

I hope this helps!

Best wishes,
Stephen



Op wo 25 mrt. 2020 om 08:43 schreef 赫雨欣 <hyxer1998 at 163.com>:

> Dear fieldtrip users:
> I just started to use fieldtrip and I have a trivial question asking for
> help.
> I’m trying to find out the labels of the sensors belonging to the
> significant clusters. I think maybe I can find some threads in
> stat.posclusterslabelmat or stat.negclusterslabelmat and in stat.mask but
>  they are all in three dimensions chan_freq_time, I don’t know how to do it
> exactly in codes.
> Hope someone can help me. Thank you very much.
> From HeYuxin
>
>
>
>
> _______________________________________________
> fieldtrip mailing list
> https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
> https://doi.org/10.1371/journal.pcbi.1002202
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20200325/cda86ce3/attachment.htm>


More information about the fieldtrip mailing list