Structure of Stieltjes classes of moment-equivalent probability laws (Q865341)

A Stieltjes class is a one-parameter family of moment-equiva\-lent (hereafter denoted M-equiv\-alent) distribution functions constructed by modulation of a given indeterminate distribution function (called the center of the class). The aim of this paper is to investigate and extend the structure properties of the Stieltjes classes of M-equivalent probability laws. Section 2 explores the simple consequences of the Stieltjes class definition, with its center, extreme elements, and distinct elements joined by the class. The paper shows that members of a Stieltjes class are mutually absolutely continuous, and an example points out the existence of many pairs of M-equivalent density functions which cannot be joined by a Stieltjes class. Any pair of M-equivalent and mutually absolutely distribution functions can be the extreme members of a Stieltjes class, but boundedness relations between members imply constraints on distribution functions which can be joined by a putative center. A distribution function is proved to be a center if it is not an extreme point of the convex set of M-equivalent distribution functions. Section 3 continues the analysis of further examples for the lognormal law. A well-known distributional identity for this law is extended to certain Stieltjes classes. Section 4 considers the problem whether two M-equivalent and infinitely divisible laws can belong to the same Stieltjes class. An author's general construction provides the substance for new examples and counter-examples. In particular, a certain M-equivalent family is exhibited and discussed as an infinitely divisible example which, subject to parameter constraints, can be joined to the lognormal class, acting as a center. Section 5 shows that new Stieltjes classes can be constructed for a given Stieltjes class by random scaling, this fact relying on an established connection between the lognormal and the \(q\)-gamma laws.