[FieldTrip] 2 x 2 with Two Between Subjects- ANOVA

[Stephen Politzer-Ahles]

Dear Paul,

Yes, based on your description, you would not be able to run a 2x2 ANOVA
with this design. I guess an alternative would be to test the two simple
effects (using independent samples t-tests) without first testing the
interaction; this isn't optimal, but at least you'd be able to make claims
about whether there is an effect in each group (although you won't be able
to compare the size of the effect across groups).

Alternatively, if I recall correctly, I think this kind of ANOVA is
implemented in another toolbox, mne-python. But I'm not sure if it's really
statistically valid, given that the Fieldtrip tutorial says it shouldn't be
possible.

Best,
Steve

---
Stephen Politzer-Ahles
The Hong Kong Polytechnic University
Department of Chinese and Bilingual Studies
http://www.mypolyuweb.hk/~sjpolit/
<http://www.nyu.edu/projects/politzer-ahles/>



> Message: 6
> Date: Wed, 10 Jul 2019 19:28:21 -0400
> From: Paul Dhami <pdhami06 at gmail.com>
> To: fieldtrip at science.ru.nl
> Subject: [FieldTrip] 2 x 2 with Two Between Subjects- ANOVA
> Message-ID:
>         <CABqnCmA4W8ZwCX7foXBYdJSfVkWOh=
> 04nPXkH1N9Zd7tcq7YZg at mail.gmail.com>
> Content-Type: text/plain; charset="utf-8"
>
> Dear Fieldtrip community,
>
> I have 4 groups of participants that fall into one of four
> categories/factors:
>
> Clinical/Young, Clinical/Old, Healthy/Young, Healthy/Old.
>
> I was hoping to test an interaction in a 2-way ANOVA design, with Age group
> and Clinical group as my two factors.
>
> Reading this link:
>
> http://www.fieldtriptoolbox.org/faq/how_can_i_test_an_interaction_effect_using_cluster-based_permutation_tests/
>
> In the first paragraph:
>
> You can use cluster-based permutation tests for some but not for all
> > interaction effects. Specifically, you can only use them for testing
> > interaction effects in factorial designs with only a single
> > between-subjects factor
>
>
> Based on the second sentence of only being possible when there is a single
> between-subjects factor, is it safe to say that in my design, I wouldn't be
> able to use cluster-based permutation tests?
>
> What would be the best way to analyze such as design than with
> cluster-based permutation tests? Would it be just a one-way ANOVA with the
> independsamplesF function?
>
> Thank you,
> Paul
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20190725/ef4f3a5c/attachment.html>