IS THE PARTIAL MANTEL TEST INADEQUATE?

A publication by Raufaste and Rousset (2001) questioned the effectiveness of the partial Mantel test, a nonparametric statistical test for association among three distance matrices. By repeated simulations, we calculated the cumulative density functions of the null-hypothesis probability of no correlation, within an explicit model of causal relationships. Results do not support the criticism: in conditions of moderate correlation between the independent matrices, the actual error rate is closely associated to the intended type-I error a.