摘要
考虑线性回归模型。这里是已知的Ｐ维向量序列，是未知的Ｐ－维向量，称为回归系数。是随机误差序列。现在我们提供一种方法剔除一些对因变量Ｙ影响总和可以忽略的变量，以使建立的模型更加稳定，并在不假定随机误差是独立同分布的条件下，给出了这种模型选择方法的强相合性的证明。
Conside the linear regression model where is a sequence of known p-vectors, is an unknown p-vectors, known as regression coefficients. {ei} is a sequence of random evrors.We propose a sort procedure,it get rid of variable exert a neglectable amount of influence on Y. To make the model more stable and more manageable compulationally. Here we do not assum that the random is i.i.d. (independent and identically distributed).Under condition of relaxation, model selection procedure is proved to be strongly consistent.
关键词
线性回归
模型选择
最优子集
强相合性
linear regression optimal subset model selection strang consistent
