Hi @fehiepsi, could you elaborate on that last comment? Are you suggesting a mistake in my approach?
I’m using AutoDelta to learn estimates for all the continuous variables in my model. AutoDelta seems to work better than AutoNormal.
Additionally, this approach of using an automatic guide to learn the remaining variables and then inferring the discrete variable separately works better than defining a manual guide which has a discrete variable included – this guide in particular. Although, the results are still not as good as they can be, so any suggestions for how I could improve things would be much appreciated.