Fighting Gerrymandering with PyMC3
summary
Bayesian hierarchical ecological inference (King‑Rosen‑Tanner style) gets implemented with PyMC3 implementation to produce precinct‑level posterior distributions rather than point guesses. Simulation recovers true minority/majority rates; applied to 2016 North Carolina it matches about 6.8M of 7.1M voting‑age people, yields 2600 precinct estimates and about two‑thirds of votes matched. Results show ~75% nonwhite→Dem vs ~35% white→Dem, enforces rates in [0,1] and reports clear posterior uncertainty.
AI-written summary · about
chapters (20)
- 0:00Intro
- 1:23PLAN FOR THE DAY
- 1:53WHERE I'M COMING FROM
- 5:03VOTING RIGHTS ACT
- 7:31SHELBY V. HOLDER (2013)
- 9:30QUESTIONS OF INTEREST - ECOLOGICAL INFERENCE
- 11:26WISHLIST
- 12:06GOODMAN'S ECOLOGICAL REGRESSION
- 14:12A HIERARCHICAL BAYESIAN APPROACH - THE MODEL
- 16:14WITH COVARIATES
- 16:40WITH MORE CATEGORIES (RXC)
- 17:01A HIERARCHICAL BAYESIAN APPROACH - SIMULATION STUDY
- 18:17SIMULATING DATA
- 20:14MODELLING!
- 21:22VISUALIZING RESULTS
- 25:19NORTH CAROLINA
- 28:21OVERALL VOTE FOR DEMOCRATIC CANDIDATES
- 28:33ESTIMATED NON-WHITE VOTE FOR DEMOCRATIC CANDIDATES
- 28:57ESTIMATED WHITE VOTE FOR DEMOCRATIC CANDIDATES
- 29:15IT'S A GREAT TIME TO CARE ABOUT GERRYMANDERING
related talks
36mClusterf*ck: A Practical Guide to Bayesian Hierarchical Modeling in PyMC3
44mAn introduction to PyMC3
39mProbablistic Programming Data Science with PyMC3
40mHierarchical Bayesian Modelling with PyMC3 and PySTAN
01h 42mRecreating, Understanding, and Visualizing FiveThirtyEight's...
01h 10mLies damned lies and statistics in Python
35mVariational Inference and Python
30mDisease Modeling with Scipy and PyMC