An Extension of Marshall and Olkin's Bivariate Exponential Distribution
本文推广了Marshall和Olkin的二元指数分布，使其绝对连续且不必无记忆，新边际分布失效率递增，联合分布呈现老化模式，并比较了极大似然和矩估计的计算效率与精度。
Abstract This article extends Marshall and Olkin's bivariate exponential distribution such that it is absolutely continuous and need not be memoryless. The new marginal distribution has an increasing failure rate, and the joint distribution exhibits an aging pattern. It offers an advantage in separately identifying the shock arrival rates and their impacts. Regarding estimation of the model, both maximum likelihood and method-of-moments-type estimation are considered. The former is more efficient but computationally more demanding, whereas the latter is simpler in computation but less efficient. The trade-off between computational burden and efficiency is gauged through Monte Carlo simulations, and it turns out to be favorable for the method-of-moments-type estimation.