[Eeglablist] NewCrossf -- bootstrap

[Arnaud Delorme]

Hi Makoto,

yes, in theory you are right. In practice though, in most cases, coherence values in EEG are shaped like an exponentially decreasing distribution. This is because, when there is noise in the two channels in which you are assessing coherence, coherence amplitude is close to 0. So using a lower threshold would not make sense in that context I think. However, in case of intracranial data where there is less noise, this might be relevant.

This might be worth investigating further. 
Maybe Tim could comment on that as well.

Thanks,

Arno

On Aug 29, 2012, at 8:30 AM, Makoto Miyakoshi wrote:

> Dear Arno,
> 
>> Coherence values are complex numbers and the function only return an upper threshold
> 
> Is it a test against mean values from a baseline period?
> Without confirming details, I remember Rodriguez et al. (1999)
> reported in Science event-related decrease of coherence ('perception's
> shadow'...) We may want to test the decrease in addition to increase.
> Can we do it just using the negative value of the calculated
> threshold?
> 
> Makoto
> 
> 2012/8/28 Arnaud Delorme <arno at ucsd.edu>:
>> Dear Michele and Makoto,
>> 
>> yes, there was a problem with the documentation. The number of values for the bootstrap thresholds is 18 values. Coherence values are complex numbers and the function only return an upper threshold (the absolute value of the coherence should be larger than a specific number). I have updated the help message of the function.
>> 
>> If you want to see the mask of significant values, you may type
>> 
>> mask  = mcoh > repmat(cohboot', [1 size(mcoh,2)]); %or more efficiently mask  = bsxfun(@gt, mcoh, cohboot');
>> figure; imagesc(mask);
>> 
>> Thanks,
>> 
>> Arno
>> 
>> On Aug 27, 2012, at 10:30 AM, Makoto Miyakoshi wrote:
>> 
>>> Dear Arno,
>>> 
>>> She says cohboot does not produce Matrix (nfreqs,2) as explained in
>>> the help but instead it produces Matrix(freq,1). Would you please help
>>> us?
>>> 
>>> Makoto
>>> 
>>>> The function help say that:
>>>> %       cohboot     = Matrix (nfreqs,2) of [lower;upper] coher signif.
>>>> limits
>>>> %                     if 'boottype' is 'trials',  (nfreqs,timesout, 2)
>>>> 
>>>> But I obtain always a Matrix(nfreqs,1) and I'm not able to use it.
>>> 
>>> 2012/8/23 Michele Barsotti <nomeserio at gmail.com>:
>>>> Hi dear experts,
>>>> 
>>>> I'm using the function newcrossf() to calculate coherence between two
>>>> electrodes.
>>>> 
>>>> As optional INPUT parameters I give the following:
>>>>        type: 'phasecoher'
>>>>      amplag: []
>>>>      subitc: 'off'
>>>>     shuffle: []
>>>>     detrend: 'off'
>>>>       rmerp: 'off'
>>>>     winsize: 256
>>>>    timesout: 895
>>>>    padratio: 1
>>>>       freqs: [3 40]
>>>>      nfreqs: 18
>>>>   freqscale: 'linear'
>>>>    baseline: [-2.5039e+003 -3.9063]
>>>>       alpha: 0.0500
>>>>       naccu: 200
>>>>    boottype: 'shuffle'
>>>>    baseboot: 1
>>>>    condboot: 'abs'
>>>>     topovec: [5 9]
>>>>       elocs: [1x13 struct]
>>>>    chaninfo: [1x1 struct]
>>>>      maxamp: 1
>>>>       title: 'coherence between, C3 C4'
>>>> 
>>>> And newcrossf returns as OUTPUT:
>>>>        mcoh: [1x18 single]
>>>>    timesout: [1x895 double]
>>>>    freqsout: [1x18 double]
>>>>     cohboot: [18x1 double]
>>>>   cohangles: [18x895 single]
>>>>    allcoher: [18x895x20 single]
>>>>      alltfX: [18x895x20 single]
>>>>      alltfY: [18x895x20 single]
>>>>         coh: [18x895 single]
>>>> 
>>>> Now, I would like to work with the coherence bootstraped but I don't know
>>>> how to use the cohboot:[18x1] with the coherence matrix coh: [18x895].
>>>> 
>>>> The function help say that:
>>>> %       cohboot     = Matrix (nfreqs,2) of [lower;upper] coher signif.
>>>> limits
>>>> %                     if 'boottype' is 'trials',  (nfreqs,timesout, 2)
>>>> 
>>>> But I obtain always a Matrix(nfreqs,1) and I'm not able to use it.
>>>> 
>>>> Any suggestion?
>>>> 
>>>> Thank you
>>>> 
>>>> Michele
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> --
>>>>       -Michele-
>>>> 
>>>> _______________________________________________
>>>> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
>>>> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu
>>>> For digest mode, send an email with the subject "set digest mime" to
>>>> eeglablist-request at sccn.ucsd.edu
>>> 
>>> 
>>> 
>>> --
>>> Makoto Miyakoshi
>>> JSPS Postdoctral Fellow for Research Abroad
>>> Swartz Center for Computational Neuroscience
>>> Institute for Neural Computation, University of California San Diego
>> 
> 
> 
> 
> -- 
> Makoto Miyakoshi
> JSPS Postdoctral Fellow for Research Abroad
> Swartz Center for Computational Neuroscience
> Institute for Neural Computation, University of California San Diego