[FieldTrip] Significance of the cross correlogram

[Simon Nougaret]

Hello everybody,
I'm a new Fildtrip User and I use it principally to compute
crosscorrelations analysis.
I tried to find a topic in the archives of the discussion about a technique
to compute significance level of the correlations performed by filedtrip
but I didn't found anything.

My first question concern the ft_spike_xcorr function where I do not
understand the case 'shiftpredictor' line 272.
Why the 'shuffling' method means to compute the cross correlation between
the trial n of the neuron 1 on the trial n-1 of the neuron 2 ?
Why is it not a real shuffle ? (to keep the same 'shuffle' curve if we redo
the analysis ?)

My second is about significance between the shuffled and real
cross-correlogram.
Do you think that it is correct, instead of the lines 275 to 278 :
          inTrial1_old = spike.trial{indx(1)}==cfg.trials(iTrial-1);
          ts1_old      = sort(spike.time{indx(1)}(inTrial1_old(:) &
inWindow1(:)));
          inTrial2_old = spike.trial{indx(2)}==cfg.trials(iTrial-1);
          ts2_old      = sort(spike.time{indx(2)}(inTrial2_old(:) &
inWindow2(:)));

to perform the analysis between trials taken by chance ?

          RandTrial = 1 + (nTrials-1) * rand(1);
          RandTrial = round(RandTrial);

          inTrial1_shuff = spike.trial{indx(1)}==cfg.trials(RandTrial);
          ts1_Rand      = sort(spike.time{indx(1)}(inTrial1_shuff(:) &
inWindow1(:)));
          inTrial2_shuff = spike.trial{indx(2)}==cfg.trials(RandTrial);
          ts2_Rand      = sort(spike.time{indx(2)}(inTrial2_shuff(:) &
inWindow2(:)));

The idea is to compute that thousand times to have the possibility to get
an equivalent of the p.value (as a kind of bootstrap analysis).
I join a figure of this computation where we can easily calculate p-values
for each bins of the crosscorrelogram. For each bin the value of the black
curve (real crosscorrelogram) will be compared to the level of chance based
of the thousands of repetitions.

Thank you in advance for your feedbacks and your responses.

Best regards,
Simon Nougaret

-- 
___________________________________________
INVESTI SUL FUTURO, FAI CRESCERE L’UNIVERSITÀ:

*DONA IL 5 PER MILLE ALLA SAPIENZA*

CODICE FISCALE *80209930587*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20160711/dc479a23/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Example shuffling.png
Type: image/png
Size: 196932 bytes
Desc: not available
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20160711/dc479a23/attachment-0001.png>