[FieldTrip] 3x2 ANOVA using permutation tests?

Stephen Politzer-Ahles stephen.politzer-ahles at ling-phil.ox.ac.uk
Tue Jul 19 11:16:06 CEST 2016

Hi Anne,

I agree with Tineke. The best thing, I think, would be to calculate the
difference waves for each individual (essentially turning it into a
1-condition design with 3 groups) and then use indepsamplesT to compare the
three groups; this is conceptually the same as testing the group*condition
interaction; see
I guess you could also test the main effect of group (by using the two
conditions' data and treating all subjects together as if they're one
group, to compare the two conditions across all subjects) or the main
effect of group (by getting the average, rather than the difference, of the
two conditions for each subject, and then comparing that across the three
groups), but I guess based on your design that the group*condition
interaction is probably the thing of primary interest.


Stephen Politzer-Ahles
University of Oxford
Language and Brain Lab
Faculty of Linguistics, Phonetics & Philology

> Message: 2
> Date: Mon, 18 Jul 2016 12:54:30 +0000
> From: "Snijders, T.M. (Tineke)" <tineke.snijders at donders.ru.nl>
> To: FieldTrip discussion list <fieldtrip at science.ru.nl>
> Subject: Re: [FieldTrip] 3x2 ANOVA using permutation tests?
> Message-ID:
>         <815A9820E75FBC4F96B7CD3A8089D11C378ED8B3 at exprd04.hosting.ru.nl>
> Content-Type: text/plain; charset="iso-8859-1"
> Hi Anne,
> To first do the analysis on separate groups, and then do an ANOVA on the
> resulting time-windows comparing the groups is double dipping.
> The simplest solution is to first do the cluster randomization test on all
> 3 groups together (i.e. treat them as 1 group), and compare the 2
> conditions (so just a depsamplesT).
> Based on this you get a time-window & electrodes that show a difference
> between conditions.
> You then extract the subject means from this time-window&electrodes
> (meaning you get one value per subject per condition), and use these in an
> ANOVA to compare the different groups, and to look at the effect of the
> behavioral regressor.
> Alternatively, you can use the difference scores (condition 2- condition
> 1) as an input for the cluster randomization, with the three groups as
> between-subject factor (indepsamplesF).
> Best,
> Tineke
> ________________________________
> From: fieldtrip-bounces at science.ru.nl [fieldtrip-bounces at science.ru.nl]
> on behalf of Anne Mickan [amickan1990 at gmail.com]
> Sent: Monday, July 18, 2016 2:22 PM
> To: fieldtrip at science.ru.nl
> Subject: [FieldTrip] 3x2 ANOVA using permutation tests?
> Dear fieldtrip community,
> I've been using cluster-based permutation tests to analysis my EEG data in
> which I have 3 groups and 2 conditions each. Ultimately I want to know
> whether groups differ from each other with respect to differences in
> conditions. In a traditional analysis I would do an ANOVA on preselected
> time-windows, but I haven't been able to figure out how to properly do that
> within the permutation test.
> So far I simply did permutation tests for each of the groups separately.
> Then on the basis of these I choose windows for an ANOVA (and a regression
> analysis with some behavioral measures I acquired). Is that OK? Or is there
> a way to do the permutation tests for all groups together and add group as
> a between-subject factor (in addition to the within-subjects factor
> condition)? And would that be better / obligatory to do (instead of what I
> did so far).
> How do I need to change the cfg.design in order to do that?
> Thanks a lot in advance!
> Best,
> Anne
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: <
> http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20160718/7f208e09/attachment-0001.html
> >
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20160719/e1225eaa/attachment-0002.html>

More information about the fieldtrip mailing list