No label defined (Q4034417)

For a non-negative random variable \(X\) with distribution function \(F(x)\), the survival function is \(\overline{F}(x)=1-F(x)\), and the mean residual life function \(e_ F(x)\int_ x^ \infty \overline{F}(u)du/\overline{F}(x)\), \(x\geq 0\). A test is proposed for \(H_ 0\): \(F(x)=F_ 0(x)\) vs. \(H_ 1\): \(e_ F(x)\geq e_{F_ 0}(x)\) for all \(x\geq 0\), for a known \(F_ 0(x)\), apparently based on the empirical distribution function. If \(G\) is another distribution function, a form of stochastic dominance is defined by saying \[ X\geq Y \qquad\text{if}\qquad \int_ x^ \infty \overline{F}(u)du\geq \int_ x^ \infty \overline{G}(u)du, \qquad x\geq 0, \] and an asymptotically distribution-free two-sample test to test this is developed. Several related dominance situations are also considered.