Developing a Weighted Sum Statistic of z-transformation Correlation Coefficients with a Blocking Variable

摘要

This article suggests a weighted sum of z-transformation of the Fisher-Yates correlation coefficients for testing the association between two variables with a third blocking variable. In Monte Carlo simulation, the power of the weighted sum of z-transformation of the Fisher-Yates correlation coefficients was compared with the weighted sum of Kendall's taus and the weighted sum of Spearman's rhos, each with the optimal choice of weights, respectively. In the bivariate logistic distribution case, the weighted sum of Spearman's rhos is preferred; otherwise, the weighted sum of z-transformation of the Fisher-Yates or the van der Waerden coefficients is more powerful for bivariate normal distribution and bivariate exponential distribution.