BrendanNyhan
Brendan Nyhan
Note: Abramowitz model has new polarization var j.mp/SbHyzx Difficult tradeoff - want to improve model, but danger of overfitting
MattDabrowski
Matt Dabrowski
@BrendanNyhan I know he's writing for the popular audience, but I'd love to see some model fit. R^2, P values, anything. He released coeffs.
@BrendanNyhan I know he's writing for the popular audience, but I'd love to see some model fit. R^2, P values, anything. He released coeffs.
DrewLinzer
Drew Linzer
@MattDabrowski @BrendanNyhan It's pretty easy to recreate his dataset. What you'll want to look at is the residual plot.
@MattDabrowski @BrendanNyhan It's pretty easy to recreate his dataset. What you'll want to look at is the residual plot.
MattDabrowski
Matt Dabrowski
@DrewLinzer @BrendanNyhan All the better to know if a model is full of crap just by a glance.
@DrewLinzer @BrendanNyhan All the better to know if a model is full of crap just by a glance.
kwcollins
Kevin Collins
@BrendanNyhan Calling it a polarization variable is a bit misleading; isn't it just an indicator for 1996 and afterwards?
@BrendanNyhan Calling it a polarization variable is a bit misleading; isn't it just an indicator for 1996 and afterwards?
kwcollins
Kevin Collins
@BrendanNyhan Ah, I see. Still looks fishy. Clinton had net pos approval in 2000 (right?), so variable takes value of -1 only in 2008?
@BrendanNyhan Ah, I see. Still looks fishy. Clinton had net pos approval in 2000 (right?), so variable takes value of -1 only in 2008?
