May 09, 2007
Poor Poll?
Larry Brown at MLB Fanhouse notices poor polling practices in the recent survey about Barry Bonds.
Posted by David Pinto at
08:45 AM
|
Cheating
|
TrackBack (0)
another case of espn making the news so it can report the news.
I think Larry's critiques are (mostly) without merit, particularly regarding potential deception on the part of the pollsters.
I'll agree completely with Andrew.
Larry Brown's critique is based on a serious misunderstanding. The poll results are weighted, not raw. That's normal for "scientific" polling. Larry's comments are based on extrapolating from the oversampling of blacks (25% of the raw sample) and assuming that the results are also raw - thus he falsely concludes that the deck is stacked relatively in favor of the "black" point of view.
Oversampling is a technique to increase sample size for a subpopulation so that the poll can achieve reasonable reliability for the result quoted for that subpopulation.
Look at the percentages overall and among blacks and whites for the question of whether Bonds used steroids:
The results are 37% yes among black fans, 76% yes among white fans, and 73% yes overall. You get 73% overall if white fans' results weigh 11 times more heavily than black fans. The poll shows that whites are 50% likelier than blacks to be baseball fans; that goes on top of the real 7:1 ratio of white to black in the population at large, leading to the weighting of 11:1 for white fans to black fans.
Finally it's not quite fair to the poll itself to judge it by a media summary of its implications; Larry Brown simply conflates the espn article and headline with the poll.