摘要
在许多情况下,对线性回归模型我们感兴趣于选择足够多的重要预测变量。本文指出了[1]中对著名的BIC准则变量选择方法强相合性证明的错误,并重新给出了一组强相合性条件。在这组条件下,我们也证明了BIC选择方法是强相合的。这组新的条件既容易验证又应用广泛。
In many situations,what one is interested in about linear regression models is to select adequate important predictive variables. In this paper,we point out a mistake in verifying strong consistency of famous BIC criterion variable selection in[1]. Also, a set of new strong consistent conditions are given. Under these conditions,BIC selection procedure is proved to be strongly consistent. These new conditions are easily verified and can be widely applied.
出处
《高校应用数学学报(A辑)》
CSCD
北大核心
1995年第1期26-33,共8页
Applied Mathematics A Journal of Chinese Universities(Ser.A)
关键词
BIC准则
变量选择
线性回归
相合性
BIC Criterion
Variable Selection
Linear Regression Model
Consistent.