Bayes factor for each predictor?


I am using brms Version 1.10.2 on Windows7.

Using lme4, I have calculated a multilevel model, in which some predictors become significant, one does not and one is only a trend (p = .08). In my field (cognitive neuroscience) reviewers often ask for sort of a follow-up with a bayesian approach.
I would like to know whether there is a way of obtaining a Bayesian measure for each predictor within a model.


PS: I am just starting to use Bayesian approaches so please forgive if I ask “stupid” questions ;).


A Bayes factor is for the model as a whole. You can use it to compare two models one with a predictor and one without, but it does not really make sense to repeat that for each predictor in the model. The key assumption of a Bayes factor is that one of the models you estimate is the true one, which is a strong assumption.

A better approach is probably to ask what is the posterior probability that a coefficient has the wrong sign. See the hypothesis function in the brms package.