Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Supported by

BF10 and BFinclusion in Bayesian ANOVA

Hi!

I have a 2 x 2 x 2 repeated measures design and would like to add Bayes Factors to the main and interaction effects of my regular ANOVA results. We are however puzzled by the BF10 & BFinclusion results of 1 of our effects (high vs low salience items). If we understood corrctly:

  • BF01 (BF10) - a comparison of our model including the effect (i.e., salience) to the null model.
  • BFinclusion - as comparison of all models that include the effect (i.e., salience) to all models that do not include the effect.

In the regular ANOVA we get a F-value of 13.8 (p < .0001). However, there is a large discrepancy between BF10 (~1.6) and BFinclusion (~460). How is this possible? Also, when computing BF10 for a simple ttest (by converting F to t) and putting the t in the online Bayes Factor calculator, the Bf ~= 40....

Further, we were wondering what is more correct to a add to regular ANOVA results? BF10 or BFinclusion?

Thank you in advance!

Nathalie


Comments

  • The differences between the BFs for "salience" might be explains as stemming from the fact the other models that also include "salience" are much better than the models that do not include "salience" (which is the definition of the inclusion BF).

    For anova tables, I've found that Inclusion BFs (for matched models) are best. Read more about matched models here>>.

Sign In or Register to comment.