Bayesian rmANOVA Model Comparison and Post-hoc tests deviate
I ran a Baysian rmANOVA with a factor time and a factor conditions. I am unsure how to interpret these results though. There seems to be string evidence for the main effect of time compared to the null model. But when I look at the post-hoc tests there doesnt seem to be evidence for an effect of time. This is consistent with the descriptive plot:
I was first thinking that it may be because the main effect of time collapses all conditions but they do not seem to be very different even then.
Can anyone help me interpret this output?
Thank you for your help!
Comments
Dear Kwellstein,
This result does look anomalous and I don't really get it immediately. I do see that you added a number of terms to the null model. I am particularly suspicious about the terms "subject" and "random slopes". I will also bring this to the attention of our expert. Sorry about the massive delay, I completely dropped the ball on this.
EJ
My understanding is that, technically, JASP does not have a Bayesian Repeated-Measures ANOVA. While, it's called that, it's really a Bayesian Linear Mixed Effects model--thus the "subject" and "random slopes" terms.
R
Yes, but they were explicitly added to the null model, which should have happened automatically
My bad, this happens by default and it's mentioned in the model specification!
Can you share the data, kwellstein?
The plot may not about the same numerical situation that's dealt with the the RM model, since it's not clear to me that the descriptives plot accounts for age or gender. A may to check this would be to see if the plot changes at all when you remove age and gender from the model. Another thing you could do is create a new variable that splits age into, say, three categories, and then create six plots with each showing a gender-by-age-category condition.
R
If the plot were to show estimated marginal means,and not just simple descriptives, that's what you'd want.
R