Correcting Doubly Truncated Correlations: An Improved Approximation for Correcting the Bivariate Normal Correlation when Truncation has Occurred on Both Variables

Abstract

Although the problem of correcting correlation coefficients that have arisen from range-restricted distributions has received a great deal of attention in the psychological testing and measurement literature, very little of that literature has dealt with the problem of restriction on both the independent (X) and dependent (1) variable. Wells and Fruchter (1970) presented an approximate correction in such situations. This article reports and improved approximation. This new approximation is recommended for all situations where restriction of range is suspected on either variable or on both X and Y.