[Mne_analysis] spatiotemporal cluster tests in EEG sensor space

[Maryam Zolfaghar]

        External Email - Use Caution        

p.s.
con = find_ch_connectivity(epochs.info, "eeg") gave me the following so I
could not use it for the connectivity parameter of the statistical
functions,

*Could not find a connectivity matrix for the data. Computing connectivity
based on Delaunay triangulations.*
*-- number of connected vertices : 129*



On Wed, Jun 19, 2019 at 9:13 AM Maryam Zolfaghar <
Maryam.Zolfaghar at colorado.edu> wrote:

> Hi all,
>
> I am new to using cluster analysis for my time-frequency EEG data. I am
> still a bit confused of how exactly "
> *mne.stats.spatio_temporal_cluster_test*" works. I first wanted to try
> setting all parameters according to the following fieldtrip page (The
> configuration settings of  ft_timelockstatistics ):
> http://www.fieldtriptoolbox.org/tutorial/cluster_permutation_timelock/
>
> However, I could not find any parameters for some of them, for example:
>
> cfg.clusteralpha    = 0.05;            % alpha level of the
> sample-specific test statistic
> cfg.clusterstatistic = 'maxsum';    % test statistic that will be
> evaluated under the                               permutation distribution.
> cfg.minnbchan       = 2;                 % minimum number of neighborhood
> channels
> cfg.clustertail         = 0;
> cfg.alpha                = 0.025;          % alpha level of the
> permutation test
>
> If I set *connectivity* to its default (None), MNE statistical function's
> outputs gave me the followings: (my input '*X*' is [cond1, cond2], each
> cond is averaged over a frequency range and with the shape: n_subj X n_time
> X n_chann)
> *t_obs* with the shape (n_times X n_chann),
> *cluster_p_value* as an one-dimensional (n_times*n_chann) array, and
> *cluster* as a connection between t_obs and cluster_p_value to find
> corresponding indices.
>
> It seems it does all pixel-wise statistics and not any cluster-based
> correction. So
> 1. Is the "threshold" the alpha level of the sample-specific test
> statistic (similar to cfg.clusteralpha in fieldtrip ft_timelockstatistics)?
> 2. How to set connectivity matrix to consider both time and spatial
> correlation? to have clusters instead of pixel-by-pixel statistical values
> (t_val)?
> 3. In fieldtrip, cfg.clusterstatistic could be set to 'maxsum', how should
> I set this in MNE?
> 4. How to get the threshold value of TFCE?
> 5. To plot the result (shape n_times X n_chann) with topoplot, it got the
> average of statistics over one channel and all time points, so when I
> create a topoplot of a boolean matrix (True for significant channels and
> False for non-significant ones), it does not give me a topoplot map with
> only two colors (e.g., red for significant channels and blue for
> non-significant ones). Is there any function to show only significant
> channels on topoplot map in a better way?
>
> Thank you,
> -Mary
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20190619/adf259bc/attachment.html