Statistical Tests

Abstract

THE discussion in these columns which has followed Mr. Buchanan-Wollaston's letter in NATURE of August 3 tempts me, as yet a third member of University College, London, to try to express my views on an interesting, but evidently controversial, subject. Others besides myself have, I know, welcomed a reminder from the author of the ” Grammar of Science” that, in the fullest sense of the terms, we are concerned not with the truth or falsehood of a hypothesis, but with the extent to which it graduates our observations of Nature. I fancy that if Mr. Buchanan-Wollaston were to consider specific examples he might find it a little difficult to substantiate the statement, made in his letter in NATURE of November 2, that hypothetical frequency distributions are of two radically different kinds. If I understand him correctly, he suggests, for example, that the Normal curve sometimes falls into one category, sometimes into the other. But can he really maintain that this distribution curve with its infinite limits for the variable is ever anything but a graduation formula ?