Mark Blumenthal writes:
HuffPost Pollster has an immediate opening for a social and data scientist to join us full time, preferably in our Washington D.C. bureau, to work on development and improvement of our poll tracking models and political forecasts.
You are someone who has:
* A passion for electoral politics,
* Advanced training in statistics and dynamic Bayesian data analysis,
* A Ph.D. in statistics, political science, economics or the social sciences or comparable high level training or experience,
* A desire to make a lasting contribution in the way the news media cover polls and elections.
We are:
* The award-winning website formerly known as Pollster.com, which joined the Huffington Post in 2010 and remains the internet’s premier source for uniquely interactive polling charts and electorate forecasts and a running daily commentary that explains, demystifies and critiques political polling.
* Home to the open source Pollster API, which provides academic researchers and software developers with one-of-a-kind programmatic access to the results of tens of thousands of opinion poll results we have collected since 2004.
* Part of the Huffington Post, the Pulitzer Prize-winning source of breaking news, features, and entertainment, as well as a highly engaged community for opinion and conversation. The Huffington Post has 50 million monthly unique visitors (comScore November, 2012) posting more than 8 million comments each month.
To apply, please send an email with cover letter and resume to [email protected] with the subject line “Pollster data scientist.”
This looks great. It’s different from our postdoc position with Yougov in that it’s more focused on the polling application, whereas our position is more about Stan and general statistical methodology. We may well do some collaboration, though.
Looks like a fun opportunity. I had a lot of thoughts about hierarchical modeling and election prediction during my obsessive poll tracking during the last two cycles, though I lacked the historical data to actually implement them. I feel there is a lot of room to make more statistically rigorous/unified models as compared to those on 538, for example. That is not to say that Nate Silver’s modeling is bad. He always asks the right questions, and uses data appropriately to get the answers (which is of course the important bit about stats).