Note on selecting the better component of the bivariate exponential distribution (Q1321762)

Let \(X\) and \(Y\) be random variables having \textit{A. W. Marshall} and \textit{I. Olkin's} [J. Am. Stat. Assoc. 62, 30-44 (1967; Zbl 0147.381)] bivariate exponential (BVE) distribution whose survival function is given by \[ \overline{F}(x,y)= P(X>x, Y>y)=\exp [-\lambda_ i x-\lambda_ 2 y- \lambda_ 0 \max(x,y)], \] where \(x>0\), \(y>0\), \(\lambda_ 1>0\), \(\lambda_ 2>0\), and \(\lambda_ 0\geq 0\). This BVE distribution is derived from supposing that failure is caused by three types of Poisson shocks on a system containing two components. The problem is to select the better component with respect to \(\lambda_ i\) \((i=1,2)\) which are viewed as the hazard of the components in lifetime analysis. When \(\lambda_ 1<\lambda_ 2\), we define that \(X\)-component is better than \(Y\)-component and we select \(X\)-component. When \(\lambda_ 1>\lambda_ 2\), \(Y\)-component is selected as the better component. Some known results for selecting the better component of BVE distribution are stated. The probability requirement and the sample size for the requirement are given. Further, we give a two-stage sampling scheme to satisfy the requirement of normal approximation.