[FieldTrip] Linearly indexed data for mvar connectivity

Antony Passaro antony.passaro at gmail.com
Wed Mar 19 16:27:18 CET 2014

Dear Jan-Mathijs,

Thank you for your reply, I appreciate it. In regards to the potential bug
in the ft_freqanalysis_mvar function, I see that it does indeed "swallow"
 linearly indexed combos, however, a reshape error is typically returned
(line 146) since it tries to reshape the matrix into the incorrect number
of elements. Perhaps I am doing something wrong, but I only feed the
untouched output from the mvaranalysis to the freqanalysis function. I will
send the "bug" to Bugzilla.

In regards to your point about DTF and PDC normalization, as I understand
it, one is normalized by the outflow and the other is normalized by the
inflow (DTF and PDC respectively?). That being the case, it seems like
a normalization to only one channel (or a subset of channels) or from only
one channel (or a subset) would produce a correct normalization in one case
(ie outflow) but not in the other (ie inflow). PDC and DTF taken together
should then show the correct inflow but not outflow (PDC?) and the correct
outflow but not inflow (DTF?), correct? I know that proposal may not be the
most mathematically viable option, however, I'm not sure how else to tackle
this issue.


On Wed, Mar 19, 2014 at 9:38 AM, jan-mathijs schoffelen <
jan.schoffelen at donders.ru.nl> wrote:

> Dear Tony,
> I don't think that these immediate plans exist. I wasn't aware of the fact
> that ft_freqanalysis_mvar does not swallow the linearly indexed
> combinations. We may want to look into that. Could you post a bug on
> bugzilla for this? The fact that it is not supported for dtf and pdc
> actually has a reason, which is related to the fact that these measures are
> normalized with respect to the total inflow/outflow for a given channel,
> where the MVAR-model explicitly should be based on the multivariate
> decomposition (i.e. where multi means all channels included). The
> functionality of ft_mvaranalysis with cfg.channelcmb fits a separate
> mvar-model for each channel pair.
> Does this make sense?
> Best wishes,
> Jan-Mathijs
> On Mar 19, 2014, at 2:15 PM, Antony Passaro wrote:
> I was wondering if there were any immediate plans to implement this
> functionality for the mvar connectivity measures (ie DTF, PDC, etc.)?
> Linearly indexed data such as specific channel pairs (instead of all
> possible channel pair combinations) seems to work fine for the
> ft_mvaranalysis function but fails for the ft_freqanalysis_mvar function
> and is not fully implemented in the ft_connectivity_dtf and _pdc functions.
> Any advice on how to proceed would be much appreciated.
> Thank you,
> -Tony
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
> Jan-Mathijs Schoffelen, MD PhD
> Donders Institute for Brain, Cognition and Behaviour,
> Centre for Cognitive Neuroimaging,
> Radboud University Nijmegen, The Netherlands
> Max Planck Institute for Psycholinguistics,
> Nijmegen, The Netherlands
> J.Schoffelen at donders.ru.nl
> Telephone: +31-24-3614793
> http://www.hettaligebrein.nl
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20140319/7b980c13/attachment.html>

More information about the fieldtrip mailing list