Lévy isometries of the space of probability distribution functions (Q536290): Difference between revisions

Let \(D(\mathbb R)\) be the space of all probability measures on \(\mathbb R\), and \(L(F,G)\) denote the Lévy distance between \(F,G\in D(\mathbb R)\). Let \(\phi:D(\mathbb R)\to D(\mathbb R)\) be a surjective Lévy isometry, i.e., it is bijective, and \(L(\phi(F),\phi(G))= L(F,G)\) for all \(F,G\in D(\mathbb R)\). The main result of the paper states that there is a constant \(c\in \mathbb R\) such that \(\phi\) is either of the form \(\phi(F)(t)=F(t+c)\), \(t\in \mathbb R\), or \(\phi(F)(t)= 1-F((-t+c)-)\). See also [\textit{L. Molnár}, ``Kolmogorov-Smirnov isometries and affine automorphisms of spaces of distribution functions'', Cent. Eur. J. Math. 9, 789--796 (2011; Zbl 1239.46022)].