JASP Bayesian Linear Regression - Model Comparison Table
Hi!
I have a question about Bayesian linear regression; specifically the model comparison table.
I have three predictors: NE, EE and contacts. The model comparison table indicates that the best model is the one that only contains NE. However, if I click "BF10" and "Compare to best model", the model comparison table indicates a BF10 of 1.872 for the second model in the table. This second model also includes contacts and the interaction between contacts and NE (i.e., NE + contacts + NE*contacts). I conclude from this that the observed data are 1.872 times more likely under the model containing NE, contacts and their interaction as predictors compared to the model that only contains NE.
My question: Why doesn't JASP indicate this model as the best model?
Cheers,
Tabea
Comments
Hi Tabea,
It may be that you are using the default correction for multiplicity, and this does not assign equal prior probability to each of the models. In the table, the "best" model is the one with the highest posterior plausibility, and this is not necessarily the one with the largest support from the data. Is this the source of the problem? You could attach a screenshot of the table so I have a better idea of what's going on.
Cheers,
E.J.
Hi EJ,
That explains it - Thanks!
Cheers,
Tabea