You can represent trends with a drift term in an AR model. I’m talking about basically any scan model that has only one return value in the inner function. The logp inference can always work those out, and they fit really fast in nutpie.
You can represent trends with a drift term in an AR model. I’m talking about basically any scan model that has only one return value in the inner function. The logp inference can always work those out, and they fit really fast in nutpie.