Estimating the Correlation Between two Variables When Their Joint Frequency Distribution is Unknown

Abstract

Four cases pertaining to estimating the correlation between variables Y and Z from various amounts of information associated with the bivariate frequency distributions of X and Y and of X and Z are considered. A computer program for estimating the correlation between variables Y and Z when the joint frequency distributions of X and Y and X and Z are known is described. The estimates obtained by using this program have been very close to the actual values of r_yz in situations where these were known.