View Single Post
Old 08-11-2016, 09:14 AM   #114
Iowa_Flames_Fan
Referee
 
Iowa_Flames_Fan's Avatar
 
Join Date: Jan 2005
Location: Over the hill
Exp:
Default

Quote:
Originally Posted by GGG View Post
http://fivethirtyeight.com/features/...ig-difference/

Some neat info on how Silvers model works and how it over reacted to the Quinnipac Pennsylvania poll and this most recent selzer poll.
Yes, it's an interesting read. I didn't realize Selzer had Clinton ahead by 18 in March--that now seems like a not-particularly-relevant data point (too far from the election, and perhaps an outlier--winning by 18 would be a truly unprecedented landslide). However, because the model looks at trend lines within a poll, it reacts strongly to a Selzer poll showing Clinton up by only 6.

I still think his model is likely very solid. I like that the "polls plus" version doesn't overreact to polls taken months before an election, and that it expects the race to tighten as the election gets closer, since they normally do.

On that note, we are still a log way from the election, but it's important to note that early voting in some swing states actually starts after Labour Day in some swing states.
Iowa_Flames_Fan is offline   Reply With Quote
The Following 3 Users Say Thank You to Iowa_Flames_Fan For This Useful Post: