The stability of a characterization of the bivariate Marshall-Olkin distribution (Q1192553)

The authors introduce a class of weakly bivariate NBU (briefly WBNBU) distributions which survivor functions possess the following two properties: \[ \begin{alignedat}{3} &G(x+t,y+t) && \leq G(x,y)G(t,t)\quad && \text{ for all } x,y,t\geq 0;\tag{i}\\ &G_ i(x+t) && \leq G_ i(x)G_ i(t)\quad && \text{ for all } x,t\geq 0,\tag{ii} \end{alignedat} \] where \(G_ i\) is the marginal survivor function, \(i=1,2\). The subclass of WBNBU where the partial derivatives with respect to \(x\) and any \(y\) exist is denoted by WBNBU*. It is shown that a bivariate c.d.f. from WBNBU* is a Marshall-Olkin (M.O.) distribution iff an expression \(\beta(G)\) written in terms of the first and second moments of the corresponding random vector, and its marginal hazard rates at the origin, equals to zero. The same \(\beta(G)\) gives an upper bound on the average distance between an arbitrary WBNBU* and the corresponding M.O. distribution. Stability of the given characterization of the M.O. distributions in the class of WBNBU* c.d.f's is established. Some other inequalities between the uniform distance and Lévy distance are also found. A short review of other works on stability of the M.O. and of the univariate NBU distributions is given.