[FieldTrip] Partial Directed Coherence threshold with few trials

[Thiago Figueiredo]

Hello Fieldtrip Community,

I was analyzing corticocortical (EEG-EEG) and corticomuscular (EEG-sEMG)
partial directed coherence with the ft_connectivityanalysis() function and
later calculating a frequency dependent threshold for the coherence values
in a session using my own implementation of the statistics described in
Schelter, Björn, et al. "*Testing for directed influences among neural
signals using partial directed coherence*." *Journal of neuroscience
methods* (2006).  I then considered in my analysis only the significant
values higher than this threshold.

The problem is that now we are applying a more strict preprocessing method
that rejects many more trials, leaving us with a
​small number of trials ​(sometimes 20 or sometimes even less than 10
trials per session, instead of the 50 or 60 that we had before). This leads
us to the problem of the validity of the coherence estimates with few
trials and to the problem of the thresholding method, which is based on a
theoretical asymptotic limit and gives incorrect results for such a small
number of trials.

So my question is twofold:

1- What is the minimum number of trials that we are allowed to use, in
order to obtain a reasonable estimate for the PDC with
ft_connectivityanalysis()
2- Do you know of any method for calculating the significance of the
coherence values to a given significance level, considering that we have a
small number of trials? (permutation tests? something else?)

Thank you

-- 
Thiago da Cruz Figueiredo

Neuroprosthetics Laboratory, Institute for Medical Psychology and
Behavioural Neurobiology - University of Tübingen, Germany

Núcleo de Inovação Tecnológica em Reabilitação (NITRE) - Universidade
Federal da Bahia, Brasil
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20160520/4a89096d/attachment-0001.html>