Can flawed statistics be a substitute for real biology?

Abstract

Serious problems can result when statistical tests are imposed a posteriori on biological data that were not collected according to an appropriate sampling scheme. Of particular concern are the cases where there are serious methodological flaws underlying the analysis, and yet a welter of statistical analyses appear to have been “thrown” at the data, as a substitute for asking meaningful biological questions. Inmany of these cases the data have been painstakingly gathered over many years, but the researcher(s) has failed to determine clear objectives, or employs poorly designed sampling procedures. Biologists who are specialists on the organisms that are the subject of such studies, often find that such analyses provide nothing, or very little, of any value biologically. Statistics are often used to bolster support for statements that are mere assertions and speculations. The objective of this paper is to highlight these problems from a longterm statistical study on subalpine cicadas (White & Sedcole 1993). By highlighting the methodological problems, I do not wish to detract from other positive aspects of this research.