Standards In Plotting For Inference & Forecasting Output

[AFg6K7h4fhy2]

Presently, Models in metaclass.py have a method called plot_posterior, which calls plot_posterior located in mcmcutils.py. The tutorial Fitting a hospital admissions-only model employs arviz cohesively throughout, and in previous discussions, MSR developers seemed to converge on using arviz abundantly.

Should Model be made to have additional plotting features, such as plot_trace or plot_ppc from arviz? If not, then can we develop some standards for what to plot after retrieving samples from our MSR models?

[damonbayer]

Agree in general that there should be support for more convenient plotting. If we go this direction, I think we should start with more support for converting model output to ArviZ InferenceData.

[AFg6K7h4fhy2]

Agree in general that there should be support for more convenient plotting. If we go this direction, I think we should start with more support for converting model output to ArviZ InferenceData.

Wrt your latter remark, I agree. I spent some time playing around with InferenceData because trace, HDI, and ppc plots are easier to use this way, but didn't get as far as I would like. Expected value for our inference / forecasting pipelines from really understanding arviz seems high. Hopefully #135 can demonstrate some arviz utilities not previously demonstrated.