DEPENDENCE ANALYSIS OF RANDOM VARIABLES, WITH HYDROLOGIC APPLICATIONS

Abstract

Instead of the usual correlation coefficient for determining the closeness and mono- tony of stochastic relationship between random variables, the indicator correlation is suggested, of numerical values easy to calculate at quantile curve points. The quantile curve is suggested for approximating the functional relationship between the two random variables. Certain optimum properties of the quantile curve are pointed out. For a monotonous functional relationship between two random variables of exponential distribution, the quantile curve representing the relationship will be shown to coincide with the line of orthogonal regression.