Foucaulf
Jr. Member
Posts: 1,050
|
|
« on: November 15, 2014, 12:39:48 PM » |
|
I was going to say that Silver didn't allow for poll comparison to lagged averages in order to deduce a causal effect, but turns out he did (the y-axis is "deviation from trailing polling average"). I guess, to make a more accurate estimate of PPP's behaviour, you can estimate the effect on a time-series of PPP polls of time-series of lagged polls from other agencies.
The illiteracy in this thread is a sight to behold, though. Everyone knows that polling data have absurdly high non-response rates and is susceptible to questionable weighing, so why does no one seem to care - and even call out people who think this is a bad thing? Maybe it's because, if herding is true, the primary good dished out by pollsters is confirmation bias, which is what Atlas posters really want.
|