Use of Stereographic Projection for Statistical Problems

Abstract

MR. NAYLOR'S recent article¹ on an approximate method for estimating multiple correlation should direct attention to the many possible uses of projection methods in statistical work. The proof that he quotes from my lengthier and more accurate procedure referred primarily to the substitution of a stereographic projection in demonstrating Pearson's equations for correcting for selection²: the special problem of partial correlation was treated as an extreme case of selection. (There is incidentally an unnoticed error in Pearson's own diagrams; his placing of D and G relative to E and F violates the conditions of consistency which must obtain for correlations between three real variables.)