EMPIRICAL-EVIDENCE WITHOUT DATA - CRITICA L COMMENTS ON A STUDY OF THE REPRESENTATIVITY OF THE ALLBUS

We examine studies of representativity as ritually conducted for socia l surveys from the point of classical statistics. Our criticism of suc h studies concentrates on the following topics: The unattainable goal of the representative sample as a miniature of the population; the unj ustified use of the chi2 goodness-of-fit-statistic to judge difference s between the sample and population; the implicit use of alternative h ypotheses in testing that are not made explicit; the non-accomodation of test levels for multiple testing; the implicit but unplausible assu mptions of homogeneity; the uncritical belief in the results of offici al statistics; the absurd attempt to proof statements on nonrespondent s without any access to relevant data. This article is a reply to a st udy of representativity conducted for the ALLBUS (Allgemeine Bevolkeru ngsumfrage der Sozialwissenschaften) that appeared in KZfSS, 44,1992, pp. 315-340.