Flexible Statistical Modeling

on YouTube ↗01h 27m

" summary

Claims that BART gives calibrated probability intervals and that pymc-bart makes it practical get tested with concrete examples: F1 lap times and GSS happiness, using MIMO notebooks and Pixie. Models use the PG BART sampler, fit around 100 trees (50–200 common), and diagnostics report occasional R̂ > 1.01 and imperfect coverage. Posterior predictive checks and PSIS-LOO are used for model comparison; fits look reasonable but uncertainty bands sometimes under- or over-cover.

AI-written summary · about

" chapters (11)

1. 0:00Flexible statistical modeling intro
2. 1:08Bayesian inference and BART
3. 2:35BART vs gradient boosting
4. 15:04Bayesian inference basics
5. 27:50BART tree structure and priors
6. 33:56PyMC model specification
7. 48:00BART for F1 lap times
8. 1:00:49Model diagnostics and checks
9. 1:11:11BART for classification
10. 1:21:00Advanced BART priors and limits
11. 1:21:22Q&A

" related talks

• 43mProbabilistic Programming in the Real WorldZach Anglin · PyData DC 2018
• 42mSolving Real-World Business Problems with Bayesian ModelingThomas Wiecki · PyData London 2022
• 42mBayesian Non-parametric Models for Data Science using PyMC3Christopher Fonnesbeck · PyCon 2018
• 01h 27mProbabilistic Python: An Introduction to Bayesian Modeling with PyMCChris Fonnesbeck · PyData London 2022
• 34mFast, beautiful and easy Bayesian modelling: Can you have it all?Rhydwyn McGuire · PyCon Australia 2015
• 30mProbabilistic modelling with ScalaMartin Mauch · Scalar Conference 2015
• 01h 23mProbabilistic Graphical Models, HMMs using PGMPYHarish Kashyap K, Ria Aggarwal · ODSC India 2018
• 01h 27mBayesian Time SeriesChris Fonnesbeck · PyData London 2025