Stochastic monotonicity and the Markov product for copulas (Q2041744): Difference between revisions

A function \(c({u_1},{u_2})\) is a density of 2-copula for joint probability distribution \(p(x{}_1,{x_2})\) if \(p({x_1},{x_2}) = c({u_1},{u_2}){p_1}({x_1}){p_2}({x_2})\), where \({u_1}({x_1}) = \int_{ - \infty }^{{x_1}} {{p_1}(s)ds} \), \({u_2}({x_2}) = \int_{ - \infty }^{{x_2}} {{p_2}(r)dr} \). A cumulative function \(C({u_1}({x_1}),{u_2}({x_2}))\) with \({u_i}\) as arguments is called a 2-copula. The paper gives two different characterizations of stochastically monotone 2-copulas using the isomorphism between 2-copulas and Markov operators. The first approach establishes a one-to-one correspondence between stochastically monotone copulas and monotonicity-preserving Markov operators. The second approach characterizes stochastically monotone copulas by their monotonicity property with respect to the Markov product. Applying this result, the authors identify all idempotent stochastically monotone copulas as ordinal sums of the independence copula Π.