RM Bayesian ANOVA : simple main effect and post hoc

PM_Stat

Hello

I'm conducting my firsts analyses with bayesian approach and I need help for a RM Bayesian ANOVA.

We're conducting A RM ANOVA with one within subject factor A (A1; A2) and one between subject factor B (B1; B2).

Can I analyze simple main effect for Bayesian ANOVA ?

If yes, how conduct it ? I don't see it in JASP. So can I conduct it with bayesian t-test ? Indeed in frequentist approach it's possible to analyze simple main effet with t-test. As I have only 2 modality in each factors simpl main effect should be sufficient.

If no, should I analyze post-hoc ? Is there a correction set up in JASP ? Because when I observed my post hoc, BF01 were the same when I computed t-test. Is that mean that post hoc computing and several t-test are same ? It would be strange no ?

Moreover, one advantage of t-test in JASP is additional possibilities (graph, robustness check,...). I'm reading Jong, Marsman, Wagenmakers (2019) about correction of multiple comparaison. Thus, can we imagine that is possible to apply a correction directly in bayesian t-test in JASP (e.g., null correction) and compute comparisons that interest me.

Thanks

PM

EJ

Hi PM,

We haven't yet implemented Bayesian simple main effects. Yes I would do a t-test. The Bayesian post-hoc correction generally takes place through an adjustment of the prior model probability (so the BFs are the same for a pre-planned and a post-hoc test -- it is through the adjustment of the prior model odds that the exploratory nature of the tests is accounted for); so if the simple main effects analysis is inspired by the data then I would apply the post-hoc correction, which we have implemented by the Westfall et al. method.

Cheers,

E.J.