Some of you may have heard that there is a theory doing the rounds that Chrysler dealers who donated to Republican candidates in 2008 are more likely to have the dealership closed. Turns out that there is no evidence for this when one compares to the odds of a randomly chosen Chrysler dealership closing. A car dealer, who is likely to be a white, middle-aged man, is therefore also more likely to donate to Republicans. Ergo, more Republican donors had their dealesrhips closed, but in roughly the fraction you'd expect given the preponderance of Republicans in this field of, er, endeavour.
However, this has not stopped someone doing a regression analysis of how donating to various candidates affects your chances of having your Chrysler dealership closed. Turns out your best strategy would've been to donate to Hillary's campaign, although the correlation is still not fantastic. A lovely discussion of the responsible and irresponsible use of statistics in such circumstances can be found here.
It even includes a link to a paper on how to employ Bayesian priors. GO THE REVEREND.
wow
ReplyDelete