Some Further Notes on the Theory of Correlation

The paper, referred to in the Note of Fr6chet, proceeds from the Regression Model as given by Yule, (1). The Note of Fr6chet adds comments on the measures r and TI, listing certain limitations which follow from theoretical developments. In this note we add some comments which follow from the Regression Model, (1). Some additional references are given. 1a It seems that Karl Pearson first used the Normal Probability Surface for two variables as a model, and defined the correlation r as the product moment of deviations of the variables from their respective means, with the standard deviations as units of measure. The two straight lines located on the plane of (X 1, X2) by the method of least squares were used by Yule as the corresponding regression model for representing r, (1),