|
Abstract:
|
A nonparametric statistic, called the roughness of concomitant ranks, is proposed for testing whether two quantitative vectors are dependent. Empirical evidence shows the new statistic is normally distributed with mean and variance are given in Theorem 1. The new testing procedure is highly computationally efficient and simple, and exhibits a competent empirical performance in simulations and two microarray data analysis. We apply the new method to deal with variable screening for high-dimensional data analysis. For low signal-to-noise ratio setting, we suggest to use data binning to increase the power of the test. Simulation results show the fine performance of the proposed method with existing screening methods.
|
