How to model asymmetric mixture

A bit new to PyMC3 and Bayesian statistical modeling.

I have a series that switches between two means—one with a lower mean and one with a greater mean. Overall, the greater one is less likely to occur than the lower one. Once the switch point occurs though, it remains in the higher distributions with a different probability of switching back to the lower.

The overall goal is to find the span of possible values as the series continues forward, so I’d like to capture both the probability of switching to the greater mean and the probability of switching back.

It’s not obvious to me how to model this behavior. It’s a bit like a gaussian mixture model, but with a stickiness and asymmetry to it—the probability of being selected for either distribution depends on what distribution the last point was it.


I’ve been really enjoying learning to use this library, so thanks to all contributors here for sharing your work!