多元统计分析线性模型主成分筛选准则研究

The study of the screening criteria of the maincompnents in a multivariate linear model

对多元统计分析线性模型中系数 β估计时 ,如果观测值矩阵出现多重共线性 ,常规的最小二乘法会因均方误差太大而失效。本文的研究表明把RMS和AIC准则用作主成分评判标准 ,可降低多重共线性 ,减少均方误差 。

s:With reference to estimation of coefficient β in the model of statistics analysis of multiviriate linear model,the regular least square method is ineffective because the mean square error is too big when the matrix of the observed value has multicollinearity.This paper shows that the method of using the RMS and AIC criterion as the judging standard of the main components can cut down the multicollinearity,reduce the mean square error,and improve the stability of estimation and the matching precision of the model