(Untitled)

[fergus8]

A really nice article (http://www.scoop.co.nz/stories/pdfs/TheUEPDv00l.pdf) explaining about the exit polling problem with the 2004 election, and how nobody has satisfactorily explained (or even unsatisfactorily explained) the discrepencies. It doesn't necessarily

Comments

[buzz_overdrive]

So, he has "uncalibrated" (unweighted?) data, no info on their method, and doesn't report the exit polls margin of error? If he did report it I may have missed it. Most polls report anywhere from +/- 3-5 points for their margin of error, and most of the differences in predicted versus tallied are 3-4% points. My research background is more on the quantitative side, but those limitations on this piece sound problematic from a quantitative perspective.

I forwarded a copy to Orion. He's got proficiencies in code/quantitative research. I'm curious to hear what he says.

[fergus8]

This particular article had a good clear explanation of the bell curves and statistical probability concepts. There have been other more recent releases that both confirm and elaborate the actual numbers (which actually had the link to this article that I've linked embedded in it), though I don't have the links available at the moment. I found them with a very routine google search though.