Bayesian linear regression - how can I only compare pre-specified models (i.e., hierarchical)
I would like to conduct a Bayesian hierarchical linear regression with 4 predictors (mean affect, affect variability, affect instability and affect inertia). I want to enter these predictors in three steps:
- Model 0 - the null model (no predictors entered)
- Model A - null model + mean affect (which I compare to Model 0)
- Model B - null model + all predictors (which I compare to Model A)
However, since JASP doesn't have a Bayesian hierarchical linear regression function but only a Bayesian linear regression function, I don't know how I can prespecify the models in which I am interested. At present, I have tried to get around this by splitting this analysis into two - first, I enter all the predictors and compare this to the null model. Then, I add mean affect to the null model and compare all predictors to this mean-adjusted null model.
The problem, however, is that JASP returns a table with a model comparison for every possible combination of my four predictors (see attached). Hence, while I would like to be comparing four models (the null, the null plus mean affect, all predictors, and all predictors with the mean in the null), I am instead given output for 16 different models.
Is there a way that I can enter my predictors in steps - first with no predictors, then add mean affect, then add all predictors - and compare the full models?
Comments
I've asked our expert!
EJ