For independent nonnegative continuous random variables X_1 ,..., X_k , Z = min (X_1 ,..., X_k) and δ_i(i=1,..., k), where δ_i = 1 if Z = X_i and δ_i = 0 otherwise, are statistically independent random variables if and only if the distribution of Xi is written as F_i(x) = 1 - exp(-p_iQ(x)) (i=1,..., k). From this characterization theorem for the random variable Xi, an unbiased estimate for failure rate of distribution function is presented using reliability data in random life testing. A saving of time in a life testing experiment by allowing random censoring is also discussed.
View full abstract