An empirical study of eight tests of partial correlation coefficients

Partial correlations can be used to statistically control for the effects of unwanted variables. Perhaps the most frequently used test of a partial correlation is the parametric F test, which requires normality of the joint distribution of observations. The possibility that this assumption may not be met in practice suggests a need for procedures that do not require normality. Unfortunately, the statistical literature provides little guidance for choosing other tests when the normality assumption is not satisfied. Several nonparametric tests of partial correlations are investigated using a computer simulation study. Recommendations are made for selecting certain tests under particular conditions.