Forecasting Australia's 2019 Election with PyMC3
summary
Builds a probabilistic pipeline that turns polls into a two‑party preferred margin then into seats using historical seat‑vote data and Bayesian inference with PyMC3 MCMC sampling. Prior over margin is wide, slope of the seat relationship is about slope ≈ 3.05, and the output is a full posterior over seat counts. Final probabilities land near 25% coalition 35% ALP ~40% minority; the Buckley’s and none forecast gives credible intervals that contain the actual result.
AI-written summary · about
chapters (21)
- 0:00Introduction
- 0:47What to expect
- 1:33Outline
- 2:19What is probabilistic forecasting
- 3:17Why would you forecast probabilistically
- 7:08Defining the problem
- 8:25Working backwards
- 10:16Uncertainty
- 11:55Output
- 12:28What is PyMC3
- 13:20Bayesian Method
- 13:53Mini Model
- 14:15Polls
- 15:42Prior
- 17:51Twoparty preferred
- 19:06Seat margin
- 19:32Line of best fit
- 20:05Probabilities
- 20:22Slope and Intercept
- 24:24Results
- 25:22Discussion
related talks
01h 27mProbabilistic Python: An Introduction to Bayesian Modeling with PyMC
01h 31mProbabilistic Programming and Bayesian Inference in Python
01h 26mProbabilistic Programming and Bayesian Computing with PyMC
01h 42mRecreating, Understanding, and Visualizing FiveThirtyEight's...
42mSolving Real-World Business Problems with Bayesian Modeling
36mClusterf*ck: A Practical Guide to Bayesian Hierarchical Modeling in PyMC3
39mProbablistic Programming Data Science with PyMC3
37mPredicting political views from text