PoliticalModeling.com: House Results

Behold! My best estimate for individual House races based on polling and, come election night, the incoming results themselves. I am fairly selective about my use of polling results for this model, using only B- and better pollsters (as ranked by 538.com, and by default, not using internal polls. My running database may be found here. Slightly more detailed instructions may be found here.

Allowable parameters include:
  • ?p[n]=[x] -- Where n=0->2, and x is a number to force the fitting parameters (incumbency advantage, swing parameter, and GCB). Default: Measured by the polls.
  • ?twindow=[days] -- Set the window for polls an approximate maximum of [days]. Default: 21 (Note: I've shortened this as we get closer to the election). I use a Gaussian weighting, so that polls "twindow" days ago are only weighted about 40% as much as those taken today. At twice that (6 weeks ago), they're only weighted about 2%, and barely at all further out from that.
  • ?sigma=[sig] -- Set the global error range (in the polling). Default: 3
If you have thoughts or questions feel free to contact me directly at drdavegoldberg@gmail.com.

R → D

hi

Enter states, districts (e.g. "PA-01"), separated by commas, or "all."