Pythonic Polling Analysis and Comments on 2016's Polling Surprises
summary
Claims that 2016 polls were catastrophically wrong give way to specifics: national aggregates mostly track the result while state-by-state estimates miss because of under-sampled groups and correlated error, turnout and weighting choices. Demonstrations use Python packages, Monte Carlo sims, the Pollster API, raking and MRP and raking approaches, plus Twitter and Google Trends signals. Surprising magnitudes include a 77,000 vote margin swing and reported 8–10% response rates; internet and social-data methods sometimes outperform traditional landline sampling.
AI-written summary · about
chapters (14)
- 0:00Pythonic Polling Analysis intro
- 1:27Why polling matters
- 3:33History of political polling
- 10:55Modern polling challenges
- 14:48How polls work
- 18:58Sampling errors and weighting
- 22:27Polling aggregation models
- 26:372016 polling failures
- 28:42Polling accuracy reassessed
- 31:31Future of polling solutions
- 36:22Summary and call to action
- 41:32Q&A: Sampling bias and Twitter
- 41:53Q&A: Post-election analysis
- 44:42Q&A: Strategic voting impact
related talks
52mUsing Python to Get Out the Vote
26mBy the Numbers: Python Community Trends in 2017/2018
23mPredicting the TripleJ Hottest 100 With Python
31mA Recent History of Python Through Stack Overflow Questions and Answers
52mPython as the multitool of data science
28mWhyPy? Understanding Trends Driving the Usage of Python
32mPyPI in the face: running jokes that PyPI download stats can play on you
45mPython useRs