I have written extensively on Bayesian analysis methods. Below are links to didactic notes, official PyMC examples, blogs posts at PyMC Labs as well as write-ups of (some) of the client work I have been involved in.
Bayesian Causal Inference
Benjamin T. Vincent and Nathaniel Forde
We wrote this book for people who fit models and are then asked what would have happened if the world had been different. It covers causal designs from propensity score weighting to structural equation models. Each is built as a Bayesian model, so its assumptions can be written down, varied and questioned. Causal inference must be robust to stress-tests and sensitivity probes.
Workshops & Training
Through InferenceWorks Ltd, I deliver practical workshops to data teams and organisations across industry and academia, on causal inference, Bayesian analytics, and marketing measurement. Workshops are available remotely or in-person.
Open Source Packages
Causal inference for quasi-experiments. DiD, synthetic control, RDD, ITS, IV, and more — with Bayesian-first estimation and full uncertainty quantification. 1,100+ GitHub stars.
GitHub · Docs · PyData Talk
Causal inference explainers
- MMMs and Pearl’s ladder of causal inference
- Causal analysis with PyMC: Answering “What If?” with the new do operator
- Interventional distributions and graph mutation with the do-operator
- Regression discontinuity design analysis
- Counterfactual inference: calculating excess deaths due to COVID-19
- What if? Causal inference through counterfactual reasoning in PyMC
- Interrupted time series
- Difference in differences
Didactic/explanatory writing
- Bayesian copula estimation: Describing correlated joint distributions
- Bayesian regression with truncated or censored data
- Truncated regression in Julia/Turing.jl
- Censored regression in Julia/Turing.jl
- Simpson’s paradox and mixed models
- Binomial regression
- Bayesian moderation analysis
- Bayesian mediation analysis
- Piecewise and spline regression in Julia
- Masters-level teaching content on Frequentist and Bayesian methods
Write-ups of client projects
I spend much of my time consulting with PyMC Labs. Some of the clients I have worked with include: Alva Labs, the Bill & Melinder Gates Foundation, Colgate-Palmolive, Gain Theory, and HelloFresh. Some of the projects I have worked on have been written up:
- Bayes is slow? Speeding up HelloFresh’s Bayesian AB tests by 60x
- Estimating parameters of a distribution from awkwardly binned data
- Bayesian inference at scale: Running A/B tests with millions of observations
- HelloFresh Media Mix Modelling project
- Causal sales analytics project