PDA

View Full Version : Elections Ringleader's First Filtered Poll Analysis Of 2012: O (49.99%), R (48.32%)


RINGLEADER
09-22-2012, 03:45 PM
So I broke out my spreadsheet that has done a pretty good job of accurately predicting the final vote tallies of the presidential candidates. It takes the existing polls, weights them on historical accuracy, builds in state leads, re-weights outliers, and spits out a number for the Dem and GOP candidates as well as others.

One slight adjustment I made this year was to increase the total "other" vote just a bit. I was under a tad (less than three-tenths of a percent) last time and I think Gary Johnson will get a slightly stronger vote this time.

I went back and looked and my final run of this program on election day 2008 was pretty close: Obama's final tally under-scored 0.13% while McCain's over-scored 0.43%. Not sure how that stands up against the almighty Nate Silver, but I'll take it.

Direckshun
09-22-2012, 04:05 PM
As RINGLEADER accuses his spreadsheet of liberal bias. :)

cosmo20002
09-22-2012, 05:06 PM
So I broke out my spreadsheet that has done a pretty good job of accurately predicting the final vote tallies of the presidential candidates. It takes the existing polls, weights them on historical accuracy, builds in state leads, re-weights outliers, and spits out a number for the Dem and GOP candidates as well as others.

One slight adjustment I made this year was to increase the total "other" vote just a bit. I was under a tad (less than three-tenths of a percent) last time and I think Gary Johnson will get a slightly stronger vote this time.

I went back and looked and my final run of this program on election day 2008 was pretty close: Obama's final tally under-scored 0.13% while McCain's over-scored 0.43%. Not sure how that stands up against the almighty Nate Silver, but I'll take it.

Guy on CP vs. Gallup, etc. I'll give it the consideration it is due.

RINGLEADER
09-22-2012, 06:25 PM
Guy on CP vs. Gallup, etc. I'll give it the consideration it is due.

Just saying - it's been right the last two times.

As for bias, the filter weights the polls based on current party ID so it is what it is.