Reactie op R. Adriaanse: de toepasbaarheid van statistische toetsen op populatiegegevens

Samenvatting

In "De toepasbaarheid van statistische toetsen of populatiegegevens" (The Applicability of Statistical Tests to Population Data) R. Adriaanse argues that statistical significance tests can & must be applied to population data, particularly in the case of small populations. Adriaanse's definitions of "independent" & "random sample" are criticized. It is argued that statistical significance tests are not applicable to population data, since, except in the case of measurement error, every difference found in a population is real (even though it may not be relevant). In the case of small populations, Adriaanse's suggestion would lead to dismissing the importance of smaller correlations. In Reactie op R. Adriaanse (Reaction to R. Adriaanse) J. Praagman points out that Adriaanse operationally defines variable independence in terms of sample aselectivity (randomness), & argues that the two concepts cannot be identified. He points out that Adriaanse's interpretation of variable independence is equivalent to the random process interpretation & subject to the same criticism, ie, that the concepts of relevance & statistical significance are inadmissably twisted. In his Antwoord aan E. Bijnen en J. Praagman (Answer to E. Bijnen and J. Praagman), Adriaanse defends his definitions of "random sample" & of "independence" (per se) as opposed to his critics' concept of statistical independence. He maintains that associative measures in smaller populations are of smaller theoretical significance than in larger populations. Using an imaginary example, he shows how a r between two independent variables can be shown to be spurious by a statistical significance test. A. Orianne.