Q: It seems to me not very different from what you really wanted. I mean don't you want to ultimately put this into production in principle, and if so, there are lots of data that you would be analyzing before going into production. Isn't that a lot like splitting data -- a new data is going to be in the future and the old data is going to be in the past?

A: It's ... I think I agree with Rob's point, I should just do this sort of pure Bayes approach. The problem is that I don't know exactly how to do that in a pure Bayesian framework.

Back