线性模型信息准则的稳健性

Robustness of Information Criteria for Variable Selection in Linear Regression Model

在不需要假定:1)被选模型包含真实模型;2)误差分布为正态分布的条件下,应用作者定义的广义K-L差异度,得到了自变量选择的广义K-L差异度准则,这个准则包含了所有信息准则作为它的特例,由此说明信息准则具有稳健性。

In references,the conduction of infomation criteria for variable selectionin linear regression model needed some strict conditions. In practice,the effection of the use of these criteria doesn' t become very bad when these strict conditions are violated. This fact tells us that the infomation criteria are robust, The infomation criteria obtained under certain mild conditions. These mild conditions mean the robustness of infomation criteria.