On the copula correlation ratio and its generalization (Q2222232)

According to Sklar's theorem for any bivariate distribution function \(F_{(X,Y)}(x,y)=\mathbb{P}(X \leqslant x, Y \leqslant y)\), there exists a copula \(C\), a probability distribution on the unit square, such that \(F_{(X,Y)}(x,y)=C(F_X(x),F_Y(y))\) with the marginal distribution functions \(F_X\) and \(F_Y\). Let \(\mathcal{C}\) denote the set of all bivariate copulas. For an arbitrary copula \(C\in\mathcal{C}\) the Spearman's rho measure is \[ \rho(C)=12\int_0^1\int_0^1C(u,v) \mathrm{d}u\mathrm{d}v-3, \] and the Kendall's tau measure is \[ \tau(C)=4\int_0^1\int_0^1C(u,v)\mathrm{d}C(u,v)-1. \] These both measures describe correlation between absolutely continuous marginal distributions \(F_X, F_Y\) of the random vector \((X,Y)\) associated with a copula \(C\). Let \((U,V)\) be a bivariate random vector following a copula \(C\), i.e. \(C(u,v)=\mathbb{P}(U\leqslant u,V\leqslant v)\). The following two quantities measure the directional regression of \(V\) on \(U\) and \(U\) on \(V\): \[ \rho_{2|1}(C)=\frac{\mathrm{Var}\{\mathbb{E}(V|U)\}}{\mathrm{Var}V},\ \ \rho_{1|2}(C)=\frac{\mathrm{Var}\{\mathbb{E}(U|V)\}}{\mathrm{Var}U}. \] The authors of the paper obtain a series of properties for coefficients \(\rho, \tau, \rho_{2|1}, \rho_{1|2}\) and for related quantities. The new formulas for these and related quantities are also derived by using the special \(*\)-product between copulas.