-
Notifications
You must be signed in to change notification settings - Fork 77
Closed
Description
Just adding a note here to investigate the parameterisation options of the Bayesian IV model.
I was exploring this fitting large IV designs with lots of data and many instruments for this ticket: #229
The Bayesian IV design just takes a very long time and doesn't always give good results on large data sets. Because CausalPy hides some of the model building complexity it's harder for the user to iteratively debug and re-parameterise. I'm wondering if there is there is anything in the model specification we can do to address this.
drbenvincent
Metadata
Metadata
Assignees
Labels
No labels