Dear Jason, thank you for the link and for the kind comments here and on other websites.
I am really glad that you understand what I am doing. Although the assumptions of the FiveThirtyEight model do appear reasonable to me, they also seem unnecessarily complex.
Also, the community of fans of that site seem not to think in terms of hypothesis testing. I think it might have to do with their fantasy baseball lineage. I'm having an interesting time figuring out how to communicate with them.
Sam Wang
My impression is that most amateurs (like myself) who get into stats for Fantasy Baseball or whatever reasons... don't understand what "over-fit" means and why it is such a disaster. For most people, the more complex the model, the better/more accurate it is... and that, is your major obstacle. It seems obvious to me that the least amount of assumptions is best, but it doesn't appear that everyone agrees. If you can make an effective case for fewer parameters, then I think you "win"... but, regardless, I'll try to get you back since I think you've got the right approach.
Regarding what people think is a good model, that's very well put.
I was hoping to avoid making the site too much of a tutorial in how to construct models without overfitting. Maybe I'll use the sidebar for this kind of inside baseball (sorry, couldn't help it).
Re Ezra Klein's blog, thanks for the help. But don't worry. That guy's a blowhole and a scoundrel, and also doesn't seem to know much math. Note that his comments are remarkably lacking in mathematical substance.
If you Google "Petey" and "banned," you'll see some fascinating discussion of him on DailyKos. His M.O. is to argue forcefully, then question his opponent's integrity. Then, when he is losing, he hauls out the ad hominem attacks.
Ezra wrote me: "Ugh -- is Petey in that thread? I apologize for that. Petey is one of my trolls -- so far as I can tell, he exists to spew fairly vicious invective at me....I wouldn't let him get to you. You know you're
doing good work when the trolls come out to play."
By the way, in case you didn't know already, probability and statistics are a major component of my research program - I publish papers in which I develop new methods. In a sense, the Meta-Analysis is one such online publication, in real time. So his latest criticism is laughable.
Sam
I'm well aware of Petey's reputation and his trollish performances... for perhaps foolish reasons, I try to engage him constructively when he isn't being a total jerk. In the period of time before he resorts to ad hominems he can actually have some interesting insights... but the fact that he reliably ends up there is probably good evidence for ignoring his existence.
I haven't yet read any of your papers, or your book, but at a glance I can tell you know what you are talking about and that your view of statistics conforms largely with mine own.... I'm just a Biomedical Engineer, but in our labs' work we strive for simple models with interpretable parameters. I want to understand what is going on, not just do a curve fit... and simple models that get a fair amount of the variance are what I'm always in favor of. Adding parameters until you get the fit you want is not. That seems like what you are going for, and thus I am a clear fan.
There's another web site that uses a similar methodology as Prof. Wang but includes a factor for systematic error in the polls that could shift the results toward one or the other candidate. Systematic errors can occur due to many factors, including an incorrect estimate of likely voters, inability to reach all potential voters, survey respondents that favor one of the candidates not always willing to answer the survey or answer it truthfully, etc.
This website is located at election-projection.net and currently shows a 91% chance of Obama winning if the election were held today.
election-projection.net also has a unique feature: an Interactive Presidential Election Probability Calculator. This allows you to input your own probability estimates for each state, and from your estimates, it runs simulations to compute the probability of each candidate winning the election and an expected distribution of electoral votes.
The Interactive Presidential Election Probability Calculator can be found at
http://election-projection.net/
i...nteractive.html
Commenting by HaloScan
