Nate Cohn

@Nate_Cohn ↗

Deleted No

Hibernated No

Last Checked Oct. 5, 2021

Created

Sat Sep 18 15:56:09 +0000 2021

Likes

Retweets

Source

Twitter Web App

View Raw Data

JSON Data

View on Twitter

Likely Available

Nate Cohn @Nate_Cohn

I could go down a 50 tweet rabbit hole thinking through whether this design holds up as a theoretical proposition. Maybe I'd decide it did by the end of the rabbit hole. I'm not going to, though, because I don't think (?) there's any material consequence. It's just odd — PolitiTweet.org

Posted Sept. 18, 2021

Preceded By

Nate Cohn @Nate_Cohn

Whatever the merits of that choice (most state telephone pollsters just go with RBS only), they try to blend the two samples in an odd way: they de-dupe the RDD frame, weight each to the full population, toss together (as oppose to voters with or without RBS coverage + p sltn) — PolitiTweet.org

Posted Sept. 18, 2021

Followed By

Nate Cohn @Nate_Cohn

The risk I wonder about: do you wind up reversing non-coverage bias and bias the sample toward the non-covered population, even if only slightly, if you dedup and weight the RBS sample to the full partisan file, and don't have a party weight on the non-RBS component — PolitiTweet.org

Posted Sept. 18, 2021