Tail asymptotics under beta random scaling (Q994321): Difference between revisions

The authors try to solve the equation \[ X @>{d}>>{=} Y.B \] for \(Y\). Here \(@>{d}>>{=}\) denotes equality in distribution, the random variables \(Y\) and \(B\) are positive, independent, and \(B\) has a beta-distribution with parameters \((\alpha, \beta)\). The distribution function of \(X\) is denoted by \(H_{\alpha,\beta}\), the distribution function of \(Y\) by \(H\). Some special cases are considered, and a recursion relation is developed to obtain \(H\) from \(H_{\alpha,\beta}\). Here use is made of the Weyl fractional-order integral operator. The relation between the tail properties of \(H\) and \(H_{\alpha,\beta}\) is investigated. There is little attention to possible existence problems; for instance, if \(Y\) has a uniform distribution, then \(X\) must be unimodal. Since all random variables are positive, one might be tempted to take logarithms and use characteristic functions. This is not tried. In the very last formula of the paper there seem to be some mistakes.