Follow
@LiberalLibrarian They'll have oversampled Rs for the R primary component of the poll, and then they'll have used some fancy algorithm to "compensate" for the oversample... which of course means adjusting the poll based on their assumption about the composition of the electorate.
Polling is only reliable in blowout races. In a polarized country, you can get basically any result by how you weight the sample.
@RationalLeft @LiberalLibrarian yup sample size and content is everything. as my former boss (a stats whiz who graduated from MIT in '68 - one of the first women to do so. She also helped keep me from failing Business Stats) says: you can make #s say anything you want.