Tuesday, October 28, 2008

Sampling Errors versus Non-sampling Errors.

Iowahawk uses the power of math! to look at how the pollsters get their "poll errors."

He then looks at the errors they don't account for.

If one or more of the above statements are true, then the formula for margin of error simplifies to: Margin of Error = Who the hell knows?

Because, in this case, so-called scientific "sampling error" is completely meaningless, because it is utterly overwhelmed by unmeasurable non-sampling error. Under these circumstances "margin of error" is a fantasy, a numeric fiction masquerading as a pseudo-scientific fact. If a poll reports it -- even if it's collected "scientifically" -- the pollster is guilty of aggravated bullshit in the first degree.

Political Polls are not scientific. Errors beyond the relatively straightforward sampling ones (IE Pull n balls out of an urn containing N where N>>n) can overwhelm the sampling error.

The worst part is that only the polls near the election itself can have their accuracy measured, and that's because there's the election votes to check.

For polls well before the day of voting... there is no way to validate them.

Using one poll to "validate" the results of another poll is like using one computer model to verify another. Unless you have some experimental data to measure against you're only comparing models.

No comments: