Howdy, Stranger!

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

Supported by

Inconsistent results between regressionBF and lmBF

Hello! I am computing Bayes Factors in R and have noticed that when I use the regressionBF function in order to compare all additive models (i.e. y ~ .), the regressionBF function does not provide me with the same Bayes Factor for a given model as using the lmBF function. The regressionBF function consistently produces higher BF than the lmBF function. I am just running simple linear regressions, no interactions or categorical variables. Moreover, the top ("best") model using regressionBF actually has a lower Bayes Factor compared to the second-best or third-best models when using the lmBF function. Why is this inconsistent and which method is most reliable?

I'd greatly appreciate any input - thank you very much!

Comments

  • I'll attend Richard Morey to your question.
    E.J.

    Thanked by 1LZ1
  • Great, thank you very much!

  • They use the same functions on the backend to compute the BF, so this is likely an issue with factor definitions in R or the like. Could you create a reproducible example?
    Thanked by 1LZ1
  • Thank you very much for getting in touch - we figured out that the problem was due to the need to remove NAs, which led to inconsistencies when regressionBF was used for a subpool of the predictors. All fixed now, but greatly appreciate the responsiveness!

    Thanked by 1EJ
Sign In or Register to comment.