New York correspondent Phillip Coorey is quoted in the Adelaide Advertiser as saying:
A Newsweek poll, taken before and after bin Laden [meaning his video] appeared, gave Mr Bush a 50% to 44% lead over Senator Kerry. But with the poll's 4% margin of error, this represented a statistical dead heat.
Say what? How can 50 be statistically the same as 44? Have pollsters or journalists redefined mathematics as we know it? By my reading a '4% error' means this poll is giving Bush 50% ± 2%, i.e. 4% of the number in question. But even allowing that what the journalist means is 50% ± 4%, there is still a difference of 6% between the raw numbers. Either that is significant, given that the data is part of the same poll and therefore directly comparable, or the entire poll is questionable.
Of course, most such polls are based on 500 to 1,000 respondents from a certain area or loose demographic, so in any responsible statistical sense they probably are questionable anyway. No doubt they create employment opportunities for door-knockers, telephone canvassers and numerically challenged journalists.
No comments:
Post a Comment