[FieldTrip] multi-taper smoothing and frequency of interest

Richard Bethlehem rb643 at medschl.cam.ac.uk
Sun Sep 4 18:20:17 CEST 2016

Dear field trippers,

Would anyone be able to offer some advice on smoothing settings used for the MTMFFT method when I want to isolate lower frequencies as well as some guidance on setting the frequency of interest. What I eventually want is just the power and crosspectra for a frequency band.

So, for example I am currently looking at the delta range (2-4Hz) and then it would seem a bit odd to use a smoothing kernel of 2Hz as it would provide very frequency specific information for that range right?

In addition, I initially just set the foilim to [2 4], but this gives me 2 datapoints that I assume just refer to the information at 2Hz and 4Hz? Thus, instead I changed it to setting the foi as a logspaced set of frequencies within the delta range. However when I run that I still only get 9 datapoints/dimensions for the frequency. Can anyone explain why it would default to 9 or what the correct settings would be to simply get the power and crosspectra for a specific frequency band (at the moment I am simply averaging over the frequency range later on anyway)?




This is the code I am using:
    cfg_freq = [];
    cfg_freq.method = 'mtmfft';
    cfg_freq.output = 'powandcsd';
    cfg_freq.channel = 1:64;
    cfg_freq.keeptrials ='yes'; %do not return an average of all trials for subsequent wpli analysis
    cfg_freq.taper = 'dpss';
    cfg_freq.tapsmofrq = 0.25;
    cfg_freq.foi = exp(linspace(log(2),log(4),20));
    [freq_data.delta] = ft_freqanalysis(cfg_freq, data_iccleaned);

And this is what I used to get some adjacency matrices for subsequent network analyses:

    cfg_conn = [];
    cfg_conn.method = 'wpli';
    conn.delta = ft_connectivityanalysis(cfg_conn, freq_data.delta);
    conn.delta = ft_checkdata(conn.delta, 'cmbrepresentation', 'full','datatype','freq');
    network_delta = squeeze(nanmean(conn.delta.wplispctrm,3));

This is resting-state EEG data that has already been pre-processed and I've segmented the continuous recording into 4-second segment to create 'trials' as I want to follow up this analysis with WPLI connectivity analysis and hence I need multiple trials (correct me if I'm wrong on that as well please, but that is probably a different thread altogether).

More information about the fieldtrip mailing list