Recreating, Understanding, and Visualizing FiveThirtyEight's...
summary
Recreates FiveThirtyEight's elections forecast step by step and names the actual pieces: polls + fundamentals + experts + noise, implements pollster ratings and Cantor k‑NN for sparse districts, fits simple regression adjustments for likely‑voter screens, runs Bayesian posteriors and posterior simulation to probability of victory across 100k trials. Claims light/classic/deluxe accuracy near 96–97%, shows how within‑race error and between‑race correlation drive uncertainty, and builds an interactive boquete dashboard using NumPy SciPy and Pandas with code and slides on GitHub.
AI-written summary · about
chapters (5)
- 0:00Introduction
- 11:53Election forecasting
- 30:29Polling methodology
- 51:50Bayesian stats and demo
- 1:21:27Q&A
related talks
15m"Red State, Blue State: Election
22mElection Night Reporting Using R & Quarto
01h 05mHow to Steal an Election
10mVisualizing voting results
57mPulling Back the Green Curtain: Data Forensics, Power BI, and Dataviz
39mKeynote 3: How did you know? Explaining Black Box Model Predictions in Python
17mSecrets of Antony Green: Calculating notional margins for electoral redistributions
22mVisualizing uncertainty with hypothetical outcomes plots