Quote:
Originally posted by Lamorak
Statistically, I'd call this trend regression towards the mean. Unless I'm mistaken. any significantly large sample will likely drift towards a median as the n grows.
|
Well, this is true under the assumption that the people that have voted so far are a random sampling.
My point is that this is a biased sample.
That said, it's still the case that there are a limited number of people, so as the sample size approaches a sizeable fraction of the overall sample, the polling results will still get more accurate.
But all that suggests is that the correct polling result is under 86%. It puts no other bounds on it; the correct result could be any number under 86%, whether that's 85% or 5%.