Ah, I see. You’re referring to the code you posted in prediction_trouble.py? Have a look at the following: prediction_troubleV2.py (1.5 KB) which is an abridged version of your code. I sample from the posterior of your model using sample_ppc (in my version of PyMC3 sample_posterior_predictive isn’t valid syntax yet), and show the standard deviation of my predictions, which is the same as the standard deviation of the observations. Are we talking about the same thing, or you meant something else?
1 Like