because he is. He's a daily kos blogger (lwnj). he uses "proprietary" numerical methods that he can hide his bias in. he's in the pockets of lw media. like I said, there are better out there.
538 Polls-Only Forecast
- Thread starter qazplm
- Start date
Find me the other analysts that have 90% accuracy rates on their projections.
Or, better yet, find me all the other analysts who have, in the last two presidential elections, gone 99/100 in state projections.
I get that you don't like Silver's projections. I even think you may well have a point about where Georgia will end up. But you can't impugn Silver as a "hack" when he projects accurately for BOTH parties.
There are analysts that are better statisticians. Drew Linzer, Sam Wang at Princeton. It goes beyond who called the states right, but whose models are more consistent and whose methods are open. For example, let's say your model called Indiana for Romney by 1% and my model called it for Obama by 15%. Which is statistically better? I called the right winner, but your margin was much much better and I would have more confidence in your model than mine. This is an elementary analysis, and there are more complicated methods, but I think you see what I'm getting at.
So, being correct is secondary. You're all about purity of methodology. That's helpful.
umm...i think you missed the point about what "correct" means in statistical terms.
I don't think so. In your example, your poll predicted correctly regardless of its inaccuracy on the margin of victory.
no...these guys are not picking a binary choice. They are giving you a probability.
If I understand the argument you're making, you're saying that from a purely statistical perspective, you have an issue with Silver and 538's approach, that they are biased in the manipulation and presentation of data. Is that fair?
The point that I (and others, I think) have been trying to make is that, while the purely statistical perspective is important, when it comes to elections, being "correct" is arguably more important than being closest to the actual result. Thus, even though his statistical approach may be lacking, Silver's ability to accurately project the outcome at a 90% or higher clip means his projections carry weight.
In other words, I don't dispute that Silver is biased (everyone is biased, including, I'm guessing you'd admit as a statistician, yourself). But his bias does not mean that his projections have NO value and it certainly does not make him a "moron." It may mean his projections have limited value, but limited value is still value.
Granted, but at the end of the day, he projects a winner for each state, which is binary. And his predictions have been remarkably accurate.
When people try to hide their bias, or pass things off as scientific without letting others confirm, it throws up a lot of red flags. None of you would invest your money with a person like that or you'd end up with someone like Madoff. I've provided alternatives and the reasons why you should be terribly suspicious not only of nate silver, but of how his data is presented (especially by qaz). Obviously you're free to feel what you want. He's done well on a couple of highly predictable elections. He's been wrong predicting european events, bombed on trump, and had a rough time in the primaries. He touts his stuff as data based journalism and yet doesn't have the background in stats to back it up. I'll stick to the pros, not the fame chasing hacks.
This guy also show GA favoring Clinton and...wait for it...he's associated with Daily Kos.
I'm still laughing at how I'm "presenting" something when all I did is post a map of polls only without commentary.
He doesn't like the obvious reality so he tries to make it about something else.
He does their election special, but he's not a blogger in the pocket of NYT or espn or whoever owns 538 now. Oh and he's also a professional.
Last edited:
Similar threads
Latest posts
-
Fox News Editorial says Harris won the debate...
- Latest: Purdue95Grad
-
-
-
