Ah, I’ve actually been working on this myself recently! At least, for the case of stationary signals (yours seem to start with a high-noise regime that then decays to a low-noise regime?). Check out the discussion over in the Stan Slack Random channel. There’s a problem with sinusoidal models generally where they have crazy multimodality in the likelihood. I think I’m close to understanding why this happens, but not really close to a solution I’m happy with.
3 Likes