On the stability of some analytic results arising in the characterization of distributions (Q1585965)

Generally speaking, equations of characteristic functions of a set of random variables \(\{X_1,X_2,\) \(\dots,X_n\}\) can be constructed based on the condition of independence of the corresponding statistics. Under this framework, Zinger's theorem shows that the random variables \(\{X_1,X_2, \dots, X_n\}\) have finite moments of all orders as long as the independence of two polynomial statistics holds. This theorem is also useful in stability problems, where equations of complex form are common, to improve the order of stability estimation. In this paper, the author elaborates on stability estimation in characterization of the normal law, by dropping previously posed constraints. Moreover, the author sets forth Zinger's theorem based on the notion of \(\varepsilon\)-network in a metric space. Subsequently in the paper the existence of moments of all orders is proven for a pair of random variables \(\{X_1, X_2\}\) whose order statistics comply with a specific \(\varepsilon\)-regression condition.