Are All Economic Hypotheses False? "In other words, we can use the results of classical hypothesis tests to in turn perform a classical test of the hypothesis that a fraction Pi of null hypotheses are true. We perform such a test, and find that we can reject at the .05 level the null hypothesis that more than about one-third of unrejected null hypotheses--more than one-eleventh of all tested hypotheses--are true. Moreover, our point estimate is that none of the unrejected nulls is true. We consider three prime contenders as explanations for this finding. First, “So what? We know that all hypotheses are false--they are only approximations”; second “So what? We all know that data mining negates the classical distribution of the t-statistic, so why be surprised that test statistics fail to conform to the distribution classical theory claims they should have?”; third “Hmm! Perhaps papers that fail to reject their null hypotheses survive referees and get published if and only if the null hypotheses they test are false.” We are sympathetic to all three views, and all three have significant implications for the way economists do and evaluate empirical research. We, however, argue that the last provides the most important explanation for our findings." LOL hahahahaha Oh man, this is perhaps the best troll of economics of all time.
Posted on: Tue, 18 Jun 2013 22:46:31 +0000
Trending Topics
Recently Viewed Topics
© 2015