抄録
A simple approximate method for calculating the correlation coefficient ρ (x, y) between two probability variables x and y, say, is given in the present paper. The essential point of this method is to divide the whole range of each variable into two classes and denote the variable x (or y) by +1 or -1 according as it is larger or smaller than its mean value. If the correlation coefficient between these new series of x and y is r(x, y), we can prove the relation (3) in section §1. By using this relation, we can calculate the required coefficient ρ(x, y) with much less numerical labour than otherwise.
