This is a very interesting blog piece by Andrew Gelman on a recently published American Sociological Review article. The discussion is rather informative too. I wonder how may of the counter intuitive, novel, "interesting" or "innovative" findings reported in our journals are the result of faulty research design? At face value the study discussed employs the rhetoric of causality, yet doesn't deal adequately with an obvious self-selection problem, conditions on an endogenous variable, and extrapolates wildly outside of the region where there is any empirical data.
One should perhaps reserve a little sympathy for the author. The spotlight just happens to have fallen on them, but it could have fallen on many others (and I don't care to consider my own past sins). It does make me wonder though why, even in a "top ranked" journal, the refereeing of quantitative articles is often so woeful. In case you think I am exaggerating, here and here are links from Carina Mood's web page to another relevant case. And you might also like to consider this piece by my colleague Kenneth Macdonald.
Posts
No comments:
Post a Comment