基于伽玛与对数正态分布假设下的广义线性模型的比较和应用

本文在假设每次损失金额的变异系数相同,且它们服从伽玛分布或对数正态分布的条件下,讨论了加法模型和乘法模型的参数估计和拟合优度检验,并应用一组实际损失数据对上述模型进行了实证比较。结果表明,对于一组特定的损失数据,对数正态分布假设下的广义线性模型可能优于伽玛分布假设下的广义线性模型。

Comparison and Application of Gamma Regression and Lognormla Regression

The paper first assumes that the loss amount of every accident has the constant coefficient of variation and follows the Gamma or Lognormal distribution,then discusses the parameter estimations and goodness-of-fit tests of these additive models and multiplicative models,at the end of the paper, these models are applied to a set of practical loss data.The result shows that generalized linear models under Lognormal assumption are superior to Gamma assumption.