non-parametric two-way repeated measures anova

Discuss the jamovi platform, possible improvements, etc.
Post Reply
MM15
Posts: 11
Joined: Mon Jul 12, 2021 3:19 pm

non-parametric two-way repeated measures anova

Post by MM15 »

Is there a non-parametric alternative for repeated measures anova with more than one factor in jamovi?
User avatar
jonathon
Posts: 2613
Joined: Fri Jan 27, 2017 10:04 am

Re: non-parametric two-way repeated measures anova

Post by jonathon »

i'm not sure that such a thing exists ... so, no.

cheers

jonathon
Bobafett
Posts: 76
Joined: Thu Jul 18, 2019 11:33 am

Re: non-parametric two-way repeated measures anova

Post by Bobafett »

Strictly speaking there is not - as you are dealing with 'ranked' as opposed to 'raw' scores then this would be problematic for the interaction of the two independent variables. I suppose you could analyse the main effects of each separate IV via the traditional non-parametric route, but you still wouldn't have the interaction effect. You could be sneaky and try to analyse what would be the post hoc tests non-parametrically (assuming the interaction would have been significant), but that would be frowned upon...
MM15
Posts: 11
Joined: Mon Jul 12, 2021 3:19 pm

Re: non-parametric two-way repeated measures anova

Post by MM15 »

Thank you both
I thought it must be possible to run a non-parametric Version of this Design but it seems like there is no alternative that can also handle an interaction between two repeated measures factors
MM15
Posts: 11
Joined: Mon Jul 12, 2021 3:19 pm

Re: non-parametric two-way repeated measures anova

Post by MM15 »

Would be great if there would be an option to use bootstrapping on repeated measures anova
User avatar
reason180
Posts: 274
Joined: Mon Jul 24, 2017 4:56 pm

Re: non-parametric two-way repeated measures anova

Post by reason180 »

Though not with jamovi, you could do a permutation-test analog of a multifactor ANOVA.

Matrix of Cell Means in a 3 by 3 design:

A1B1, A2B1, A3B1
A1B2, A2B2, A3B2
A1B3, A2B3, A3B3


The test would address the question:

Given many random permutations of the data (i.e., random reassignments of data points to condition labels), and given a statistic such as 'the across-row variance of the across-column variances, how often is the magnitude of the obtained statistic at least as large as the same statistic computed on the un-permuted data?
Post Reply