On reconstruction of distributions by the translated moments of linear statistics (Q1365737)

The aim of this article is the derivation of the distribution function \(F(\cdot)\) of a random variable \(Y\) provided that its even (or odd) translated moment is known. The even translated moment is defined by the function \(\Psi(u)= E|Y-u|^q\) \((u\in {\mathfrak R})\), while the odd translated moment is defined by the function \(\Phi(u)= E|Y-u|^q \operatorname {Sign} (u-Y)\). Assuming that \(0<q<1\), the author has proved that if the function \(\Psi(u)\) is the even translated moment function of a random variable \(Y\) with distribution function \(F\), then it satisfies the following properties: (i) it is continuous on \({\mathfrak R}\), (ii) it tends to infinity as \(O(|u|^q)\), (iii) \(\lim_{u\to\infty} \Psi(u+v)- \Psi(u)= 0\), (iv) there exists a bounded and monotone function \[ \Theta(x)= \int_0^{+\infty} [\Psi(x+t)- \Psi(x-t)]/ t^{1+q} dt, \] (v) the distribution function of \(Y\) is given by \[ F(x)= 2^{-1} (1+ C_q^{-1} \Theta(x)), \quad\text{where } C_q>0. \] Moreover, if any function \(\Psi(.)\) satisfies the properties (i)--(iii), and there exists a monotone and bounded function \(\Theta\) defined as above, then \[ \int_{-\infty}^{+\infty} |x-u|^q d\Theta(x)= A\Psi(u)+ B, \] where \(A>0\) and \(B\) are some constants. Similar conclusions can be drawn for the odd translated moments. Finally, as the author points out, all the previously presented results can be extended to multidimensional analogues of the odd or even translated moments.