Doesn't any model that yields probabilities have to model those rare events (to understand polling error prob. for calibration)?
-
-
.
@zeynep Well if it can't be wrong, then it can't be right, either, and we're both wasting our time talking about it. -
.
@cwyers :-). It's an interesting philosophical question, how to think probabilistically for binary outcome rare events you can't redo. - Show replies
New conversation -
-
-
To compare two models, look at the probability that they assign to the actual outcome (the full electoral map with margins).
-
As you suggest, the 538 model assigns higher probabilities to Trump wins and Clinton blowouts than others. But they pay a price for this.
- Show replies
New conversation -
-
-
Yeah but this is just a problem for 538 bc they're hedging so much. Not really true of PEC/HuffPo/DK, even NYT.
@cwyersThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
I suspect where we'll find 538 is actually wrong is that the predicted winner wins more often than should happen by chance.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
votes aren't a random distribution. Unless you know something I don't?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Now, I believe campaigns are in a much better place to model. They have large tracking polls, etc.https://twitter.com/jmartnyt/status/795396151092674561 …
-
@jmartNYT I prefer my own method; gauging how snippy Kellyanne Conway is that day. (she was very snippy today)
End of conversation
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.