Originally Posted by Botnst

First lesson in statistics: A complete enumeration of a variable is always more accurate than any sampling scheme.

There are two major sources of statistical error: Accepting a false result believing it is true and rejecting a true report, believing it is false. These are often referred to as "Type I" and "Type II" errors. Neither of these errors are possible in complete enumeration.

Also, the possibility of unethical and inadvertent manipulation is far greater with statistical sampling. This is why blind and double-blind methods are used in high-risk/high-reward (like medicine) analyses. Even riskier are the time-dependent, bootstrapping methods employed in exit polling. Look-up bootstrapping and you'll understand what I am getting at. It is an extremely valuable and well-understood method (technically, by statisticians). Properly conducted it is very reliable.

I'm not saying that statistics are necessarily incorrect when there is a discrepancy with an enumerative poll. I'm just saying that the first thing a prudent person should do is freeze all of the data and poll results to avoid any tampering of anything. Then look first at the statistical technique and method before undermining faith in the people's vote.

If the stats stand, then look elsewhere. If stats fail scrutiny, stop there and accept the election. Undermining the voter's faith in the election process is a terrible thing to do. One's side may win this election but one may find the opponents using the same tactics in a subsequent election. That's a hell of a way to run a democracy.

Bot