Empirical likelihood based interval estimation of quantile correlation
Quantile correlation is an asymmetric correlation coefficient describing the linear relationships between two random variables.It plays a vital role in feature screening problems in many subjects such as Statistics,Finance,and Chemistry.Besides,empirical likelihood is a famous non-parametric method that is widely used in the statistical inference of several models.In this paper,the estimation equation is firstly established from the definition of the quantile correlation of any two random variables,then two proposed methods,plug-in empirical likelihood(PEL)and its adjusted version(APEL),are introduced to make interval estimation with establishing asymptotic scaled chi-squared distribution and standard chi-squared distribution,respectively.Simulation studies compare these two empirical likelihood-based methods with other methods in terms of coverage probability,interval length,and the average loss based on interval score.A real data set from Forbes magazine is adopted in the application to illustrate the presented methods.
万方数据
维普
