Journal of Data Science

Abstract: Bivariate data analysis plays a key role in several areas where the variables of interest are obtained in a paired form, leading to the con sideration of possible association measures between them. In most cases, it is common to use known statistics measures such as Pearson correlation, Kendall’s and Spearman’s coefficients. However, these statistics measures may not represent the real correlation or structure of dependence between the variables. Fisher and Switzer (1985) proposed a rank-based graphical tool, the so called chi-plot, which, in conjunction with its Monte Carlo based confidence interval can help detect the presence of association in a random sample from a continuous bivariate distribution. In this article we construct the asymptotic confidence interval for the chi-plot. Via a Monte Carlo simulation study we discovery the coverage probabilities of the asymptotic and the Monte Carlo based confidence intervals are similar. A immediate advantage of the asymptotic confidence interval over the Monte Carlo based one is that it is computationally less expensive providing choices of any confidence level. Moreover, it can be implemented straightforwardly in the existing statistical softwares. The chi-plot approach is illustrated in on the average intelligence and atheism rates across nations data.