[FieldTrip] cfg

Schoffelen, J.M. (Jan Mathijs) jan.schoffelen at fcdonders.ru.nl
Tue Dec 30 18:34:17 CET 2014

Dear Steve,
Great that you guys are interested in our HCP data! Although indeed the processed data are represented as fieldtrip-structures I think the most appropriate platform would be to ask this question on the HCP-users platform. You can subscribe to the users list on: http://www.humanconnectome.org/contact/#subscribe

For rMEG, the data.time structure is 1x147 cell, where each cell is a 1x1018 double, each going from 0 to 1.9995 (0.001966075273231 increments).

Indeed, this is the consequence of us manually chopping up the 3-minute or so recording into 2 second epochs, and removing some blatantly bad segments of data. Note that the time axis is arbitrary and does not have a meaningful relationship to a physical experimental event.
As well, we get several warnings:
“the input is raw data with 244 channels and 144 trials”, which it is not (it is preprocessed data).
“the trial definition in the configuration is inconsistent with the actual data
“reconstructing sampleinfo by assuming that the trials are consecutive segments of a continuous recording”

Don’t worry about these specific warnings at the moment

By using different cfg inputs to account for differences in the time sampling for the resting and working memory data, we get similar errors, though both output .freq structures are ~1x400 in this case.

%%for resting state cfg_rs
cfg_rs.foi        = 1:0.01:200;
cfg_rs.toi        = -1.5:0.002:2.5;

%%for wm data cfg_wm
cfg_wm.foi        = 1:0.01:100;
cfg_wm.toi        = 0:0.0019:2;

Any advice on appropriate configuration structures for wavelet analysis of this data would be greatly appreciated.

At first glance the frequency resolution requested is far too high. Given the finite and discrete sampling, not all requested frequencies can be exactly estimated. Fieldtrip will do its best, but will round off the requested frequencies to the nearest estimable one. A more meaningful cfg.foi would be 1:1:100 or so. I suggest to read up a bit about spectral analysis on the tutorial section of our Fieldtrip website.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20141230/06394813/attachment.html>

More information about the fieldtrip mailing list