PSYB10H3 Lecture Notes - Lecture 10: Sat, Procrastination, Personality Psychology
Document Summary
Polling: most polls were off, even though many models claimed a 99% confidence interval. Turnout: differences in turnout i. e. less people showed up in 2008 than in 2016, typically, when turnout is lower the republicans win. Polls: clinton +3. 2: polli(cid:374)g (cid:449)as (cid:373)istake(cid:374); is the polli(cid:374)g e(cid:396)(cid:396)o(cid:396) u(cid:374)usual o(cid:396) (cid:272)o(cid:374)siste(cid:374)t (cid:449)ith (cid:449)hat (cid:449)e"(cid:448)e seen in the past, polls are always a sample of a population. Absolute errors bounce around quite a bit: average is 2. 0. Not the case that on a national scale, polls were nationally wrong . Trump outperformed by more than the national error in key states: he won normally blue states; the polls were especially off in these states. Not going to get a perfect representation of the population in the sample: quantifiable. If you average multiple pols together, the random polls will cancel out: cancels out across polls. Can take a guess as to what systematic error might be: does not necessarily cancel out.
