Ampere is exploiting a wide range of different packages to do inference, many of which have their own routines for generating summary or diagnostic plots, and some don't, which results in a diverse range of visual styles. We should try to make sure that the plots are as consistent as possible, but without writing our own plotting routines unless absolutely necessary. Arviz provides a good way to do this for many routines, but involves us figuring out how to convert outputs to InferenceData objects, and we also have to understand how to make plots from other codes look similar.
But the big prerequisite is figuring out how we want our output plots to look in the first place! Suggestions are welcome :-) Please post examples here if you have a good idea. @sundarjhu @ciskakemper @SaschaZeegers @jontymarshall
Ampere is exploiting a wide range of different packages to do inference, many of which have their own routines for generating summary or diagnostic plots, and some don't, which results in a diverse range of visual styles. We should try to make sure that the plots are as consistent as possible, but without writing our own plotting routines unless absolutely necessary. Arviz provides a good way to do this for many routines, but involves us figuring out how to convert outputs to InferenceData objects, and we also have to understand how to make plots from other codes look similar.
But the big prerequisite is figuring out how we want our output plots to look in the first place! Suggestions are welcome :-) Please post examples here if you have a good idea. @sundarjhu @ciskakemper @SaschaZeegers @jontymarshall