Hi,
I had a look at the Betancourt’s tutorial and I’ve tried to plot the pair plots between the different parameters I’m estimating (I’ve updated the colab notebook above).
This is an example for one subject, focusing on the decay parameter
I cannot see anything obviously wrong but maybe I’m missing something.
I also looked at the likelihood space and it looked ok.
Is there anything else that I could check to debug this model, or is it just misspecified?
Also, does the scale of the parameters have an effect on the quality of the sampling?
Beta is [0,20ish], while alpha and decay in [0,1], could that be an issue or MCMC samplers are ok at handling that?
Thanks!