ABSTRACT
The paper compares several methods for testing the hypothesis that two independent groups have identical correlation coefficients. Here we consider not only Pearson's correlation, but several robust analogs as well. The main goal is to find some method that performs well in simulations, in terms of Type I error probabilities, when sample sizes are fairly small. Several methods have been proposed in the literature, but they are known to be unsatisfactory. Various bootstrap methods were considered in this study, but only one was found to perform reasonably well.
Keywords: