摘要
本文研究了两种ZIP模型,比较了一般ZIP回归模型和ZIP(τ)模型对相同的零膨胀(zero-inflated)计数数据的拟合效果,发现前者在拟合时有较好的稳定性,并在此基础上得出实际数据中两种模型应用的技巧以及需要注意的问题。同时研究了ZIP模型在保险费率厘定中的应用,将这两种模型分别应用于保险数据,发现在拟合和预测两方面,ZIP回归模型都明显优于ZIP(τ)模型。
Two kinds of ZIP models are studied in this paper. Goodness-of-fit for zero-inflated count data of common ZIP regression model and ZIP(τ) model are compared, seeing that the ZIP regression model gets a better result. Based on this, we draw some conclusions on applications of the two models. The two models are applied to ratemaking. It is shown that the ZIP regression model gets better results in both fitting and forecasting.
出处
《数理统计与管理》
CSSCI
北大核心
2013年第5期854-862,共9页
Journal of Applied Statistics and Management
基金
国家社会科学基金项目资助(08BTJ001)