My experiment is a between-subject repeated design. (Treatment-Control)
My hypothesis testing is around comparing slopes of my d.v.And null hyp being that my control has a higher slope value than treatment. (~roughly along this line)
My goal is to show the different Bayesian approaches namely (Parameter Estimation, Bayes Factor and Cross-validation) for my hypothesis testing.
Could someone guide me with a rough pipeline?
I did, dig around the forum and blogs but still, a lot of things feel foggy when writing my pipeline.
1 Like
First, you fit 2 models: 1. One where you only have one slope parameter and 2. one where you allow the slope parameter to vary by group.
To calculate a Bayes factor that compares theses models, you can use the bridge sampling package https://cran.r-project.org/web/packages/bridgesampling/index.html.
For the parameter estimation approach you can report mean and credible intervals for the difference in slopes between groups, which you can get from model 2 above.
For cross validation based comparison of the models, you can use the loo package. See e.g. this post: https://datascienceplus.com/k-fold-cross-validation-in-stan/ (including comments)
4 Likes