An empirical study of the accuracy of corrections for restriction in range due to explicit selection

Abstract

An empirical study of the corrections for restriction in range due to explicit selection resulted in the following conclusions. (1) The corrected sample correlation was no more accurate than the uncorrected sample correlation for low unrestricted population correlations in the range .10 to .25. (2) For large unrestricted population correlations in the range .60 to .80, the corrected sample correlation was always more accurate than the uncorrected sample correlation. (3) For moderate (.30 to .55) unrestricted population correlations, the corrected sample correlation was typically more accurate than the uncorrected sample correlation. (4) The correction was very sensitive to moderate departures from linearity but was quite robust in the face of rather substantial departures from homoscedasticity.