基于多重共线性的处理方法.pdfVIP

30 2 Vol. 30 No. 2 2010 6 MATH EMAT ICAL THEORY AND APPLICATIONS Jun. 2010 * (中南大学数学科学与计算技术学院, 长沙, 4100 5) 多重共线性简称共线性是多元线性回归分析中一 个重要问 消除共线性的危害一直是回归分析 的一 个重点目前处理严重共线性的常用方法有以下几种: 岭回归主成分回归逐步回归偏最小二乘法 Lasso回归等本文就这几种方法进行比较分析, 介绍它们的优缺点, 通过实例分析以便于选择合适的方法处 理共线性 岭回归 主成分回归 逐步回归 偏最小二乘法 Lasso回归 Based on M ultiple Collinearity P rocessingM ethod M an Jingluan YangW ei ( School ofM athem atics Science and Computing Technology, CSU, Changsha, 4 100 5) A bstrac t M ulticollinearity referred to as collinearity is a mu lti- linear regression analysis in a very difficult issue. H ow to elmi inate the collinearity hazards regression analysis has been a priority. The literatu re at hom e and abroad to deal w ith serious collinearitym ethods commonly used are the follow ing: R idge regress ion, principal component regres- sion, stepw ise regress ion, partial least squares m ethod, Lasso regression. In th is paper, a com parative analysis of thesem ethods and describe their advan tages and d isadvantages, easy to select the appropriatew ays to deal w ith co-l linearity through the examp le analysis. K eyw ords Ridge regress ion Principal componen t regress ionm ethod Partial least squares regression Lasso re- gression , R. Frish, 1 k X ( i = 1, 2, ,, k ) : KX + KX + ,KX + E = 0, i 1 1 2 2 k k K K ,K , : K, E I 1 2 k 0, ] * : 2009 8 2 106 E= 0, ; Ey