BEGIN:VCALENDAR
VERSION:2.0
X-WR-CALNAME;VALUE=TEXT:Optimizing State Legislative Election Forecasts - Carl Klarner
PRODID:-//Harvard events data//EN
BEGIN:VEVENT
UID:event_1428936_0
SUMMARY:Optimizing State Legislative Election Forecasts - Carl Klarner
DESCRIPTION:Please join us on November 18th from 3-4:30pm in CGIS K354 for Carl Klarner’s talk on "Optimizing State Legislative Election Forecasts"All are welcome to attend.Abstract:State legislative elections offer a unique opportunity to test a variety of methodological issues that give insight into forecasting and elections generally. This is because of the large number of such elections. How to best predict state legislative elections is also important in itself. The efficacy of competing modeling decisions was assessed with a variety of tests. These entailed “drop one analyses” in which one biennium of elections were dropped in turn from a sample of elections spanning 1968 to 2012. Next “pseudo predictions” were conducted for state legislative elections conducted in 48 states between 2007 and 2012. Last, true out of sample predictions for the 2014 state legislative elections—publicized beforehand—were also conducted. The competing modeling decisions included those that dealt with 1) scale (use of chamber versus district level models), 2) functional form of the dependent phenomena, 3) complexity, including use of the “unit rule” for modeling national level factors, 4) case selection, 5) missing data, and 6) assessing uncertainty.Please reach out to Carl Klarner to request a draft of the associated paper.
LOCATION:K354
STATUS:CONFIRMED
DTSTART:20141118T200000Z
DTEND:20141118T213000Z
END:VEVENT
END:VCALENDAR