[FieldTrip] FT clusterplot

Schoffelen, J.M. (Jan Mathijs) jan.schoffelen at donders.ru.nl
Wed May 4 11:08:05 CEST 2016

Dear Hweeling,

I am not a big fan of this function, and I would strongly recommend not to talk about ‘significant clusters’: using cluster-based permutation tests allows for inference about the difference across conditions/groups, based on a heuristic involving the estimation of clusters from the thresholded observed (and permuted) data, yet the clusters themselves are just a byproduct of this inferential procedure.

>From what I can read in the code (which you also had a look at), indeed at some point a squeeze operation is performed in case the input data contains both a time and freq axis. Yet, note that the squeeze operation only has an effect if the freq dimension (the second dimension in the matrix) is singleton, i.e. if there’s only a single frequency bin in the data.
In other words, even if the function does not crash when inputting time-frequency data, it will most likely produce ‘incorrect’ visualizations, or at least you think you are looking at something that will be different from the thing that you are actually looking at. The only way around this would be to do  a pre-selection of a certain frequency range prior to calling ft_clusterplot.
This preselection can be achieved with ft_selectdata in general (use cfg.frequency, and cfg.avgoverfreq), but will not likely produce the required output for the ‘posclusterslabelmat’ and ‘negclusterslabelmat’ fields, that are required for ft_clusterplot. At least, not if you select multiple frequency bins. Thus, in summary, you can pull this off, if you first use ft_selectdata, but in that case ensure that you select a single frequency bin.


> On 02 May 2016, at 12:06, Hwee Ling Lee <hweeling.lee at gmail.com> wrote:
> Dear all,
> I hope this email finds you well.
> I had a statistics results that I would like to use ft_clusterplot to plot the significant clusters. However, I'm experiencing some difficulties to specific the configuration.
> From the ft_clusterplot.m, I noticed that the function is able to squeeze the 3D matrix into a 2D matrix for plotting purposes. My current statistics results is in "chan-freq-time" dimension, and I was wondering how I could specific the configuration to select a specific frequency to plot.
> Can someone please help?
> Or do I have to adjust the statistics results variable manually such that it has a "chan-time" dimension for plotting?
> Thanks!
> Cheers,
> Hweeling
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip

More information about the fieldtrip mailing list