muon2
Moderators
Atlas Icon
Posts: 16,821
|
|
« on: September 14, 2011, 06:05:31 PM » |
|
Nate has part of statistical analysis wrong on the keys. He bases his statistics by saying that the keys have predicted 38 out of 38 elections. In fact the keys were inferred from 31 of those elections - from 1860 to 1980. The model has since been applied to only seven elections so its track record is not as statistically anomalous as Silver thinks. You can't count as successful predictions the input data to the model.
A second error is that the model as presented in the book makes clear that it does not attempt to predict the margin of victory. Since the variance was as large as Silver shows, the model showed no promise to make that prediction. Since the authors don't claim it, why attack it?
A third complaint (but so much from Silver) is the subjective nature of the keys. The book is far more specific with how they should be judged based on the 31 election cycles used to construct the model. For instance, the short term economy key will be set by the direction of the economy after the first quarter of 2012. Social unrest would require unrest such as in 1968, or other earlier equivalent outbursts. A third party challenge much produce a candidate that can garner 5% of the popular vote. Etc. Yes, they are subjective, but not as much as one might think just reading the bullet points.
|