Stochastic Models with Memory for Seismic Risk Evaluation
提出两种带记忆的随机模型（线性递增风险率模型和连续时间马尔可夫链模型）来替代传统无记忆泊松过程，利用秘鲁利马的地震数据验证了前者拟合更优，并探讨了其对保险和再保险的意义。
Most seismic risk studies assume the occurrences of large earthquakes follow a memoryless Poisson Process. Yet, current scientific knowledge describes earthquake occurrences as a memory process of gradual accumulation and sudden release of energy. After reviewing some suggested models for earthquake occurrence, two stochastic models with memory, the Linearly Increasing Hazard Rate (LIHR) model and a continuous time Markov chain model, are offered as simpler alternatives. Using earthquake occurrence data from Lima, Peru, the LIHR model is tested against the Poisson and found to provide a significantly better fit. Some implications of the use of memory processes for insurance and reinsurance problems are explored.