Nate Silver

@NateSilver538 ↗

Deleted No

Hibernated Yes

Last Checked March 19, 2020

Created

Sat Jul 13 18:04:54 +0000 2019

Likes

Retweets

Source

Twitter for Android

View Raw Data

JSON Data

View on Twitter

Likely Available

Nate Silver @NateSilver538

@BrendanNyhan Most people are pretty bad at building models if we're being honest so I don't really care about the consensus. Models that don't conform to the underlying structure of the data are a lot more fragile, especially in edge cases. — PolitiTweet.org

Posted July 13, 2019 Hibernated

Preceded By

Nate Silver @NateSilver538

@BrendanNyhan Ehh, I kind of think this is wrong. If the underlying system your model describes is bounded between 0 and 1, your model probably should be also. — PolitiTweet.org

Posted July 13, 2019 Hibernated

Followed By

Nate Silver @NateSilver538

@BrendanNyhan IMO if you're not in a predictive context that makes it *worse*. Models are supposed to describe reality and linearity is a lazy shortcut when you're modeling a nonlinear system. Maybe in prediction you can "get away with it" if your predictions hold up OK, but not for inference. — PolitiTweet.org

Posted July 13, 2019 Hibernated