Inclusion probability plot vs. BF(inclusion)?
Hello JASP fellows!
I'm building a bunch of linear regression models where the inclusion probabilities show that only some variables exceed the prior inclusion line. But then this doesn't always fit with the BF inclusion. Sometimes the BF inclusion is still smaller than 3, even if they exceed the line or the ones that do not meet the line don't have a BF inclusion <0.3, i.e. supporting exclusion. Which one should I rely on to build a better model in the next step?
Comments
Can you give a concrete example?
Hi EJ,
I'm trying to attach a photograph. Here you see the model-averaged coefficients and their BF inclusion as well as the prior inclusion plot. I'd be very grateful to know which one of them to rely on to improve my models.
Hi eniseg2,
I am not sure about the discrepancy -- when the prior inclusion plot suggests a small increase in probability, the inclusion BF suggests weak evidence in favor of adding that component. There is nothing special about a BF value of 3. If the BF is 2.5, there is still evidence favoring inclusion, and the prior inclusion plot will show this. Let me know if I'm missing the point!
Cheers,
E.J.