As you probably remember from my frequent recent mentions, I've been working my way through a book on probability, statistics, and the philosophy of science, STATISTICAL INFERENCE AS SEVERE TESTING, by Deborah Mayo.
This will be my last quote from that book. From p. 367, "It is supposed in many fields of social and biological science that nearly everything is related to everything: 'all nulls are false.' Meehl dubbed this the crud factor."
What follows is a description of a psychology survey done by the University of Minnesota in 1966 in which all of the following showed some degree of (often low though statistically significant) correlation -- attitudes toward school, leisure activities, educational plans, parents' occupations, siblings, birth order, family attitudes. religious preferences, MCAT scores, etc. Every null hypothesis is false.
Mayo tells us that Meehl here is also eager to explain that this is not what statisticians call a "Type 1 error". It is not that the study mistakenly found correlations where they weren't there or would disappear on closer analysis. They were and are in fact there. Everything in society is in fact mixed up with everything else, and that commingling shows up in large data bases. You can make a number of claims from such data, some of them plausible others not.
But the simple point is: beware the crud factor.
Comments
Post a Comment