On the subexponential property of a class of random variables (Q5951089): Difference between revisions

Let \(X\) and \(Y\) be independent nonnegative random variables, with the distribution density functions \(F_X\) and \(F_Y\), respectively. The class \(S\) of subexponential distributions contains those (distribution) functions \(F\) whose tail \(\overline F=1-F\) satisfies the relation: \[ \lim_{t\to\infty} {F*F(t) \over F(t)}=2,\tag{1} \] where ``*'' denotes the Stieltjes convolution. A (distribution) function is majorable at infinity if the following inequality holds: \[ \varlimsup_{t\to\infty} {\overline F(t)\over F(2t)} <\infty, \tag{2} \] the class of these functions being denoted by \(D\). Let \(V\) denote the distribution function of the random variable \(\min(X,Y)\) of the above considered independent nonnegative random variables \(X\) and \(Y\). A recent result of A. L. Yakymiv (1997) shows that if \(F_X\), \(F_Y\in D\cap S\), then \(V\in S\). The aim of the present mathematical note is to generalize this result. Let \(R(t)=-\ln \overline F(t)\), \(t\geq 0\). The main theorem of the article proves that if \(R_X\) and \(R_Y\) are boundedly semiadditive functions and \(F_X\), \(F_Y\in S\), then \(V\in S\). The proof of the theorem is based on the following established facts: (a) If \(F_X\in D\), then \(R_X\) is boundedly semiadditive. (b) (Rogozin, 1996) If \(F_X\in S\), then \(R_X\) is an asymptotically semiadditive function.