DÄ internationalArchive23/2010Differences Should Be Clarified

The authors are to be congratulated for expressing criticism of technical errors in handling significance tests. However, one aspect remained somewhat neglected: the decisive criterion for relevance is not significance but effect. How often do authors use “significant” in the sense of highly or extremely relevant, or even with star markings as for expensive brandy, and less knowledgeable readers are taken in. Sometimes it is even difficult to find the difference that was tested. It is advisable to always ask for the amount of difference in tests for differences and to relate these to the resulting mean. If the number of cases is not high enough, differences of 5% are “highly significant.” The same is true for the correlation coefficient: if r=0.30, 118 cases are enough to fall below the probability of an error of 0.001, but the coefficient of determination, r2 (the effect), is only 9% in such a scenario. During the subsequent interpretation of the results it will become essential to reflect more on the relevance—and less on the significance—of that effect.

DOI: 10.3238/arztebl.2010.0417b

Prof. Dr. med. H.-V. Ulmer

Curt-Goetz-Str. 93

55127 Mainz-Drais, Germany

ulmer@uni-mainz.de 

1.
Victor A, Elsäßer A, Hommel G, Blettner M: Judging a plethora of p-values: How to contend with the problem of multiple testing — Part 10 of a series on evaluation of scientific publications [Wie bewertet man die p-Wert-Flut? Hinweise zum Umgang mit dem multiplen Testen – Teil 10 der Serie zur Bewertung wissenschaftlicher
Publikationen]. Dtsch Arztebl Int 2010; 107(4): 50–6.
VOLLTEXT
1.Victor A, Elsäßer A, Hommel G, Blettner M: Judging a plethora of p-values: How to contend with the problem of multiple testing — Part 10 of a series on evaluation of scientific publications [Wie bewertet man die p-Wert-Flut? Hinweise zum Umgang mit dem multiplen Testen – Teil 10 der Serie zur Bewertung wissenschaftlicher
Publikationen]. Dtsch Arztebl Int 2010; 107(4): 50–6.
VOLLTEXT