Tuesday, October 14, 2008

Election '08

Results
Elections Canada
Wikipedia
SFU
Nodice
Fringe Parties


Ad Watch
The Winners
Bad Actors Love Harper (CPC)
See the Softer Side of Jack (NDP)
Jack Attack (NDP)
Green Shift (Liberal)
Viva Loss Vegas (CPC)
Harpernomics (Liberal)
Jack Attaque (NDP)
Pot Pourri
Bushwacked (Liberal)
Runaway Train (Greens)
Doodling Dippers (NDP)
Subtle Micro-Targeting (CPC)


Projections and Predictions
Throughout the campaign, I used public polling data to project seat totals. The first set of projections, explaining the methodology, can be found here, with my final projections here. Other updates can be found here, here, and here. So how did I do? It was a missed bag. The NDP and Bloc projections were spot on but the final Liberal and Conservative totals were outside the margin of error. Still, when compared to the other projections and predictions being made, the model held up respectably:

1. Calgary Grit predictions (total miss: 16 seats)
2. Ekos predictions (total miss: 18 seats)
3. Calgary Grit simulation model (total miss: 23.6 seats)
4. Barry Kay projections (total miss: 24 seats)
5. Andrew Coyne predictions (total miss: 26 seats)
6. UBC Stock market (total miss: 30 seats)
6. David Aiken predictions (total miss: 30 seats)
8. Democratic Space (total miss: 36 seats)
8. Kady O'Malley predictions (total miss: 36 seats)
10. Election Prediction Project (total miss: 38 seats)


Week in Review
Week 1 in Review
Week 2 in Review
Week 3 in Review
Week 4 in Review
Week 5 in Review


Posts
Debate Drinking Game
French Debate Live Blog
English Debate Live Blog
Better Know a Riding - Papineau
Election Day

Labels: , , , , , ,

4 Comments:

  • Perhaps the prediction and simulation model should also generate a list of close ridings where a 5% swing in the popular vote (likely outside of the MOE) would change results.

    While the first predictions would still stand, you could produce a result for the alternate 1 time out of 20 with the other.

    You could also add a system to weight turnout likelyhoods for different parties (since there seemed to be publicly available data on that too).

    You could run a model then that showed if there was a divergence in results according to turnout (ie: does a lower turnout increase the proportion of tory vote) while predicting results based on a normal distribution of turnout lieklyhoods outside of the last observed turnout.

    By Blogger Kyle G. Olsen, at 10:15 PM  

  • I have to say, despite our political differences, I really enjoyed your coverage of the campaign. It was a must-read from start to finish.

    By OpenID Devin Johnston, at 12:06 AM  

  • great idea! nice sharing ..

    By Anonymous Data recovery software, at 2:39 AM  

  • Good post, thanks for share and nice information, cerita seks or you can go at cerita sex

    By Anonymous Cerita Seks, at 10:30 AM  

Post a Comment

Links to this post:

Create a Link

<< Home