Bayesian RM ANOVA: Significant main effect with non-significant post-hoc effects

chhhim

Hello everyone,

I'd have another question related to the results of a repeated-measures ANOVA. In that I consider two factors, environment (3 levels) and time of measurement (2 levels).

For environment, by BFincl is 0.337. This is weak evidence for the absence of an effect.

The post-hoc test, however, does indicate evidence of equivalence across all three groups:

I struggle with explaining this issue to my readers... With a frequentist ANOVA this should not be possible, however, we are collecting evidence for H0 here instead of H1.

Can anyone help me understanding why this occurs? And maybe state whether I may interpret this as evidence for equivalence across all three groups...?

Thanks a lot!

Best Regards,

Chantal

EJ

Hi Chantal,

The difference is not pronounced, right? You would draw the same conclusion -- there is weak evidence that the groups are the same and hence the variable does not contribute. The post-hoc tests are just t-tests, so they will give results that differ from the ANOVA.

Cheers,

E.J.