A Sample of Spurious Correlation

Abstract

IN my former letter in NATURE of January 6 I attempted to prove that the arrangements of dots in a band would occur even if the numbers of which the sums were taken were entirely independent of one another, in which case a forecast regarding one of the numbers could not possibly be made from knowledge of the remainder. A forecast could only be made if it were shown that the width of the band were smaller than would be expected on the hypothesis of pure chance, and this vital point has received no consideration.