A Sample of Spurious Correlation

Abstract

THOUGH regretfully unable to do justice to the mathematical reasoning of Dr. G. T. Walker in NATURE of January 6, I may, perhaps, be allowed to say that it is of the essence of the method that those dots (each expressing a comparison of two sums of thirty items) tend to arrangement in a straight band, or strip, with fairly defined borders. It is expected that future dots will generally come within those limits; but to affirm this in a given case, to say, e.g., that the next dot will not be below a certain level, is it not, necessarily, to say something quite definite as to the character of the coming season, as that its rainfall, frost days, or other feature considered, will not be below a certain numerical value? If the one statement is warranted, so (by the nature of the case) is the other. Thus the essential point seems to me to be whether the past distribution of those dots affords a reasonable clue to their future distribution, and I do not see that my critic throws doubt on this.