[Eeglablist] How to access non-parametric statistics (cluster method Oostenveld and Maris)
Benedikt Ehinger
behinger at uos.de
Tue Jun 25 23:54:22 PDT 2013
Hi EEGlablist,
I had the same problem and found no other solution than modifying the
function statcondfieldtrip directly. As this is a low-level function you
need to prepare your input a little. It is sometime ago that I did the
simple modifications (but it was in v11.x), therefore I'm not sure a) if
it still works b) if these were the only modifications.
what I think I did to get the fieldtripstat:
Simply add to line 224 (after or before eeglab saves the relevant
stats.X values,e.g. pvals = stats.prob etc.):
fullStat = stat;
and put fullStat in the output to the function.
function [ori_vals, df, pvals,fullStat ] = statcondfieldtrip( data,
varargin );
As I wanted paired testing I used the following command:
[stats,df,pvals,ftStat] =
statcondfieldtrip(dataCell,'paired','on','method','montecarlo','correctm','cluster','naccu',1000);
where dataCell{1} = (ElecxTimesxTrials), dataCell{2} = (ElecxTimesxTrials)
But if you only need the t/F-values or p-values you can use this
function directly without modifications (they are saved in stats/pvals).
If you have trouble using the functions you are welcome to ask me anytime.
Best,
Benedikt
PS: I just noticed, that you need to put the fullStat = stat in other
places as well, if you test more than one electrode at a time (two-way
Anova) because there are "returns" in the code.
Am 26.06.2013 05:44, schrieb Makoto Miyakoshi:
> Dear Mikolaj and Felix,
>
> I haven't used the function, so I don't know. I believe the ERP plots
> are generated on the fly and the results are not saved. If you want to
> access the output, you have to identify which function generates the
> plots.
>
> Makoto
>
> 2013/6/23 Miko?aj Magnuski <imponderabilion at gmail.com
> <mailto:imponderabilion at gmail.com>>
>
> Dear Felix,
>
> I had a similar problem previously and could not find the cluster
> statistics data anywhere in EEGlab structures (perhaps I was
> not looking carefully enough - Makoto, is FieldTrip output saved
> anywhere in the workspace when doing cluster statistics from
> EEGlab?).
>
> When doing cluster statistics EEGlab uses another toolbox -
> FieldTrip. If you are unable to dig up the output of this procedure
> anywhere in EEGlab I suggest you should perform it in FieldTrip
> directly. It's a little bit harder than EEGlab because you don't have
> a GUI in FieldTrip, but all the functions are described very well and
> some really helpful tutorials are present on the FieldTrip web page
> (the page is here: http://fieldtrip.fcdonders.nl/).
> There is a function called eeglab2fieldtrip that can transform your
> EEG structure from EEGlab to FieldTrip style. All the following
> steps can be done according to:
> http://fieldtrip.fcdonders.nl/tutorial/cluster_permutation_freq
>
> I have some functions that could help you along the way
> (or even automatise the whole procedure), you can con-
> tact me if you decide to do cluster statistics directly in
> FieldTrip.
> Nevertheless, the FieldTrip output should be present somewhere
> in the workspace or at least be possible to catch in a script (if not
> it is still possible to modify one of the EEGlab functions to
> return or
> eval in your base workspace the information about cluster statistics).
>
> Mikolaj
>
> Pozdrawiam,
> Miko?aj Magnuski
>
>
> 2013/6/19 Felix Ball <Felix.Ball at hu-berlin.de
> <mailto:Felix.Ball at hu-berlin.de>>
>
> Dear list,
>
> Thank you in advance for your help. I currently try to use the
> STUDY function and especially the non-parametric test (Maris
> and Oostenveld cluster statistic). I have an experiment with
> two conditions and n subjects. What I want to know is at which
> time points and at which electrodes the two conditions differ
> from each other. The difference between conditions is
> investigated on the grand average level.
> I read about a method introduced by Maris and Oostenveld and
> it seems to be perfect for my needs and it is already
> implemented in EEGLab. What I get with EEGLab (see matlab
> code) is a plot (see figure C1vsC2) with different time
> intervals marked in grey for each electrode. These grey shaded
> areas are, as far as I understood, significant differences
> between conditions. Furthermore, they are in a time range
> where I expected to find a significant difference between
> conditions. Now I have 2 questions:
>
> 1) I want to access the statistics that are the basis of this
> plot. Does someone know where I can find them or do I have
> bugs in my code, preventing me from accessing the stats
> variables? I was not able to find any variable in my workspace
> resembling the results shown in figure C1vsC2.
>
> 2) The result I expected (after reading the Maris and
> Oostenveld paper), slightly differs from the result I receive.
> I expected only one cluster of neighbor electrodes and one
> continuous time interval to be shaded in grey in the plot.
> This cluster- time interval combination would be the maximum
> effect between conditions. However, looking at figure C1vsC2,
> I see that neighbor electrodes have a grey shaded areas, but
> these areas are not necessarily continuous and are not equal
> for all electrodes. If I just look at one electrode, e.g. FC3
> (see appendix), I see that the time range is sometimes
> interrupted. How do I know when conditions start and stop to
> be significantly different, if there is not only one time
> interval of significant difference? I need to know the time
> range because I want to pick the mean amplitude (in the time
> range) for a regression analysis.
>
> Thank you for your help and kind regards. felix
>
>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> <mailto: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
> <mailto:eeglablist-request at sccn.ucsd.edu>
>
>
>
>
>
> --
> Makoto Miyakoshi
> Swartz Center for Computational Neuroscience
> Institute for Neural Computation, University of California San Diego
>
>
> _______________________________________________
> 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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20130626/56fdf447/attachment.html>
More information about the eeglablist
mailing list