Um, poll Twitter. State errors may be correlated. More interesting Q: can we model a rare event based only on 11 previous cases—as 538 does?
-
-
.
@HerbieLewis The whole debate people are having is about 538 probability distributions; little debate on what those actually are. -
They can at least place upper bounds on the prob of rare events with few inputs, no? Albeit not very low upper bounds.
End of conversation
New conversation -
-
-
You can't do it well, but you have to try if you want to report a win %. Otherwise you're sweeping it under the rug, no?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
What does 'very few cases' mean? They have thousands of polls across 1000s of elections, depending on defn. of election
-
.
@DinoStraciatela@HerbieLewis 538 model is based on 11 previous presidential elections. That's it. It's a tough thing to model rare events. - Show replies
New conversation -
-
-
I believe W. S. Gosset was using two and Fisher n = 8. What's important is the polls close gap on probability sample space.
-
and by that I mean do stratified sampling really really well. Match actual voting base has to be super hard. And can't check votes!
End of conversation
New conversation -
-
-
they simulate individual states, not just national so it's 11*50...
-
But the total model is past presidential elections.
- Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.