2x2 repeated measures Bayesian ANOVA (interaction)

Ajestudillo

Hi all,

One reviewer asked me to include bayesian analysis in my paper, but I am not sure that I am getting this right, so any help would be much appreciated. I paste the output with BF10 and with BF01:

I think my analysis shows the following:

1. The differences in face conditions are 1.063e +14 more favoured than the lack of differences b (BF10 = 1.063e +14, posterior M = 5.47, SD = .55, CI = 4.29 – 6.54). 
2. The lack of differences in cond was 4.88 times more favoured compared to the differences between groups (BF01 = 4.88, posterior M = -.25, SD = .92, CI = -2.23 – 1.5).
3. The lack of interaction was 4.14 times more favoured than the interaction (BF01 = 4.14, posterior lM = -.28, SD = .51, CI = -1.32 – 0.72).

Is this interpretation correct? My understanding is that for the interaction I need to check the analysis of the effects table, am I right?

Any help would be much appreciated.

Thanks!

JASPlearner

Hello! I have a relevent question about the model comparsion results in bayesian repeated ANOVA.

Let's say that the model I am interested in is the gender+pitch+interaction mode. The BF10 of my interested model is way larger than the null model(8.864X10^38) but it is smaller than the best model (8.192x10^39). Can I say that my interested model is also supported by the data? If I say that, how can I explain the effect of the best model? If I can't say that, is that means the best model in bayesian results is the only model that is supported by the data?

EJ

Dear Ajestudillo,

1 & 2: yes

3: yes, (I assume you refer to the 4.009); note that you can add the main effects as nuisance factors to the null model (under the "model" tab) so you get the comparison between the two-main-effects model and the full model.

E.J.

EJ

Dear JASPLearner,

I would say that the data offer evidence against including the interaction. You may want to tick "compare to best model", or add the main effects to the null model in order to focus on the evidence for adding the interaction term.

Cheers,

E.J.

Ajestudillo

Thanks a lot @EJ !

Ajestudillo