Guys, looking at the format of my probability distributions normal(a, bN and sigma) and lognormal(bLN) in the figure below, should I make any new adjustments to my multivariate model? I await your suggestions!
It’s hard to say from just that plot, without knowing the model, so I won’t comment.
You might get some mileage from using the various diagnostic summaries and plots in arviz though, inc checking parameter posteriors plot_posterior, parameter summaries summary, sampler energy plot_energy, and posterior predictive checks plot_ppc etc
Many of the notebooks in the examples gallery demonstrate model diagnostics, inc this one
I would suggest plotting the autocorrelation function calling pm.plot_autocorr(trace) … Should converge asymptotically to zero.
You can also “perform the same study” as you did adding more samples, check if the parameters change.
Finally, you can “force a change in the prior” and evaluate if the posterior converged to the same shape as in the first case.
Thank you for the answer!!
Many thanks for the reply! The prediction of values were from the set were good!