Add uniformity check to SBC pipeline

[jhrcook]

Add the uniformity check spelled out in the original simulation based calibration paper (cited below). Keep the current diagnostics and checks in the report, as they provide additional useful information, though care should be taken when analyzing the “accuracy” of the models on simulated data.

Talts, Sean, Michael Betancourt, Daniel Simpson, Aki Vehtari, and Andrew Gelman. 2018. “Validating Bayesian Inference Algorithms with Simulation-Based Calibration.” arXiv [stat.ME]. arXiv. http://arxiv.org/abs/1804.06788.

[jhrcook]

While I am at it, it would be useful to collate the MCMC diagnositcs that I recently added to the reports and show plots relating the overall status of the MCMC process. For example, what is the distribution of the number of divergences per chain and variability in the number of divergences per simulation.

[jhrcook]

Sources:

• "Validating Bayesian Inference Algorithms with Simulation-Based Calibration” (original paper)
• Betancourt’s post on the Bayesian workflow: Principled Bayesian Workflow - Computational Faithfulness
• Colin Carroll’s example implementation in PyMC3: post, GitHub
• Stan implementation: ‘rstan’ function, ‘rstan’ vignette
• Andrew Gelman post
• R package ‘sbcrs’ for SBC with ‘rstan'

[jhrcook]

I have added the uniformity check to the simulation_based_calibration_helpers module. 🥳

A few more changes:

• [x] Add a step to the SBC pipeline such that each model uses the same mock data. The rule should create the mock data and save it in the cache. I can add this feature to the CLI for SBC as another typer command.
• [x] For the report, my first quick-and-dirty visualization method was to randomly select 10 variables to show. Change this to show each variable, and 3 for multi-dimensional variables (i.e. varying effect variables).