为帮助精算师在不同的数据环境下选择最优的准备金评估方法,美国非寿险精算师协会组织开发了一个产生模拟索赔数据的开源软件系统——损失模拟模型,然而损失模拟模型是否能够按指定参数要求产生模拟数据需要进行检验。文章采用不同的参数估计方法和拟合优度检验方法对模拟索赔次数的分布、索赔额的趋势以及不同险种索赔次数之间的相关结构进行了统计检验,结果表明损失模拟模型对索赔次数的分布、索赔额的趋势能够产生一致的模拟,而对索赔次数数据之间相关结构的模拟存在不稳定性。
The CAS(Casualty Actuarial Society) has developed a open-source software-LSM(Loss Simulation Model)-which produces claims and transaction data to help the actuaries to select the optimal method of loss reserving with different data,but it has to be tested whether the simulation data is produced in accordance with the specified parameters and distributions.We have tested the distribution of claims number,the trend of severity and the correlation structure of claims number between different lines used different methods of parameters estimation and goodness of fit in this paper.The results showed LSM could produce consistent simulations with the distribution of claims number and the severity trend,but it is not robust in simulation of correlation structure of claims numbers between different lines.