j reacts to the idea that clinton is a lock on 270 ev, but less certain to win

see, now we're very close to a point of agreement. i'd just question the idea that the margins cited are nearly as meaningful as suggested (especially between ohio and pennsylvania, which have often voted differently). that's the same question of demographics being predictive or not, which, remember, i largely reject.

and, i'd have to wonder aloud how anybody could agree that it is almost certain that she's going to win at least 270 electoral votes, and yet much less certain that she's going to win the election. because if that isn't a contradiction (it is), it's at least cognitive dissonance.

http://fivethirtyeight.com/features/clintons-leading-in-exactly-the-states-she-needs-to-win/

(independent random variables should be very familiar not just to statisticians, but to anybody with a remotely mathematical background - computer scientists, physicists, engineers. you may not have even ventured past them, if you're less theoretical and more practical.

you couldn't do this in canada - and if you could, it wouldn't really mean anything. but, given that there is ample data for states in the american election, and the unit of the state is paramount, i might suggest that the way to model elections is to assign each state as an independent random variable. this assumes that data in ohio and pennsylvania (and in any other two states) is entirely independent. and, i think this is correct.

this isn't exotic. it's textbook. it's just a different assumption. but, if anybody knows of anybody doing modelling like this, i'd like to see it.

https://www.probabilitycourse.com/chapter3/3_1_4_independent_random_var.php)

(i'm just trying to tersely explain what an independent random variable is for random onlookers.)