Reply to Peter Mueser

Autor: Donald N. McCloskey
Rok vydání: 1990
Předmět:
Zdroj: The American Sociologist. 21:26-28
ISSN: 1936-4784
0003-1232
DOI: 10.1007/bf02691778
Popis: On statistical significance our differences are small; one might say insignificant. I agree that the importance of the "Iowa effect" depends on whether we can do something about it and what our theoretical context might be. That is what I said. The further notion that an economic theory "applies" to the murder rate if the death penalty shows up in even a small way seems to beg the question of what constitutes "applicability." The main point, after all, is that applicability is decided in the conversations of scientists not in a table of Students-i. For "samples com monly used in empirical work" in economics an insignificantly small effect by some (still unarticulated) standard might sometimes turn out to be statistically insignificant also. But samples in sociology, for example, or educational psychol ogy are often enormously larger than the quarterly national income figures since 1946 beloved of economists. With large samples the silliness of statistical signif icance cannot be evaded. Anyway, the case Mueser makes is not general. Some people think that statis tical significance is a good "first hurdle." I would say here that they are wrong. That using statistical significance sometimes works out all right is not a good enough argument to justify the present practice, universal in economics and embarrassingly common in other social sciences, of taking statistical significance as the only meaning of significance. That the drunk's keys sometimes are near the lamppost does not justify his plan of always confining a search to the neighbor hood with the best light. That a mainly irrelevant procedure generated painlessly by our canned computer programs sometimes produces the correct decision, by accident, is not much of the argument. I stick therefore to my remarks about statistical significance. Statistical signifi cance is useless, virtually, and the social science done with its help over the past thirty years needs to be done over again. We in economics have been barking up
Databáze: OpenAIRE