On measurement of stochastic dependence in multivariate data (Q5944783)

The authors review and discuss several concepts of generalized measures of dependence. These are derived from the mutual information between the joint distribution and the product of the marginal distributions, which is defined by \[ D=\int \cdots\int f(x_1,\dots, x_p)\log(f(x_1, \dots, x_p) /f_1(x_1) \cdots f_p(x_p)) dx_1,\dots, dx_p \] for a \(p\)-variate distribution with joint density \(f(x_1,\dots,x_p)\) and marginal densities \(f_1(x_1), \dots,\) \(f_p(x_i)\). An explicit expression for \(D\) is obtained for the case of a multivariate normal distribution. An application to a case study in pattern recognition is provided.