Saturday, November 20, 2004
Nice new statistics blog
Andrew Gelman has a statistical inference blog for his research group at Columbia. A useful resource, and an interesting idea for promoting a group of grad students.
Even if you're not a statistics person, this post will be of interest. It critiques a recent working paper by some Berkeley scholars (who purport to find unexplained effects of evoting technology in Florida, even controlling for past votes); turns out this finding rests entirely on two outliers, both of them large Democratic counties. A great reminder to always look at the data, even if you're going to throw it into a "fancy" regression (and especially if that regression isn't robust!). For bivariate relationships, the scatterplot has yet to be outdone.
Even if you're not a statistics person, this post will be of interest. It critiques a recent working paper by some Berkeley scholars (who purport to find unexplained effects of evoting technology in Florida, even controlling for past votes); turns out this finding rests entirely on two outliers, both of them large Democratic counties. A great reminder to always look at the data, even if you're going to throw it into a "fancy" regression (and especially if that regression isn't robust!). For bivariate relationships, the scatterplot has yet to be outdone.