About

This notebook is an introduction to InferenceData objects and their role in ArviZ. It aims to be as hands-on as possible, using examples to cover different use cases of the data in each group, examples on how to combine several InferenceData objects and so on.

The key idea behind InferenceData objects is to centralize and store all data relevant to a specific Bayesian inference run; from observed_data to predictions going through prior and sample_stats. The goal is therefore to both ease exploration and visualization of Bayesian inference results and to ease their sharing. Hence, ArviZ also provides several converter functions to transform results from common inference libraries such as PyMC3, PyStan or Pyro to InferenceData. In the future, we'll dedicate one post to each converter function, but before, we have to make sure the conversion process is worth it!