Working Papers, University of California, Department of Economics 12,6
D. N. McCloskey and Stephen Ziliak have criticized economists and others for confounding statistical and substantive significance, and for committing the logical error of the transposed conditional. In doing so they sometimes misinterpret the function of significance tests. Nonetheless, economists sometimes make both of these errors - but not nearly as often as Ziliak and McCloskey claim. They also argue - incorrectly - that the existence of an effect, which is what significance tests are about, is not a scientific question. Their complaint that in testing significance economists often do not take the loss function into account is unfounded. But they are right in arguing that confidence intervals should be presented more frequently.