February 12, 2013 · 6:30 PM
Thanks to O'Reilly for once again sponsoring our meeting and for Intuit for hosting it.
6:30 PM Networking and Pizza
7:00 PM: Announcements
7:10 PM Drew Linzer to talk:
A big story of the last presidential election was the "victory of the quants" -- a triumph of data over intuition. While traditional pundits portrayed the race as neck-and-neck, pollsters and polling analysts confidently and accurately predicted that President Obama would win reelection. How was this done?
Drew will talk about the dynamic Bayesian forecasting model he used at votamatic.org to call the outcome of all fifty states, as early as June. Combining the results of nearly 1,000 pre-election polls -- representing over 700,000 individual interviews -- the site offered real-time opinion tracking, state- and national-level forecasting, and commentary on the race. After the election, Drew's forecasts were widely recognized as among the best of the campaign.
The Votamatic project was implemented entirely using R. Drew's presentation will show how it was done: from web scraping to data handling, estimation of the model in WinBUGS, model validation, visualization, and the design and architecture of the website.
Drew Linzer is an Assistant Professor of Political Science at Emory University, and a[masked] Visiting Assistant Professor at Stanford University. He received a Ph.D. from UCLA in 2008. A statistician and expert in public opinion, Dr. Linzer is also co-author of the R package poLCA for latent class analysis. Follow him on Twitter @DrewLinzer.
For more on polling aggregation and election forecasting: