多重共线性参考文献
检验:
方差膨胀因子( VIF)
Tolerance
Condition Number
Multicollinearity and imprecise estimation．
处理方法：
岭回归( ridge regression,) 、主成分回归( PCR) 、逐步回归、偏最小二乘法( PLS) 、倾向评分法( Propensity score)、Lasso 回归、数据分组处理( group method of data handling，GMDH)
参考文献(并不全)：
Tolerance and condition in regression copmputations．
Diagnosing and dealing with multicollinearity．
Multicollinearity: a bayesian interpretation．
Principle Components Regression in Exploratory Statistical Research．
Multicollinearity in regression analysis: the problem revisited．
Collinearity and least squares regression．
Modeling data tables by principal component and PLS: class patterns and quantitative predictive relations．
The central role of the propensity score in observational studies for causal effects．
Collinearity diagnosis for a relative risk regression analysis: an application to assesment of diet cancer relationship in epidemiological studies．
Multicollinearity and imprecise estimation．
Ridge regression: biased estmiation for nonorthogonal problems．
Multicollinearity caused by specification errors．
The fallacy of differencing to reduce multicollinearity．
Propensity score methods for bias reduction in the comparison of a treatment to a nonrandomized control group．
觉得最好的办法应该是先进行自变量的选择，而不是一上来就回归。
另外这个问题似乎经济学更感兴趣，感觉医学反倒兴趣不是特别大。没有仔细研究过这个问题，隐约觉得lasso可能是最简单的。反正近来的趋势就是lasso这类方法什么都可以往上靠。
方差膨胀因子( VIF)
Tolerance
Condition Number
Multicollinearity and imprecise estimation．
处理方法：
岭回归( ridge regression,) 、主成分回归( PCR) 、逐步回归、偏最小二乘法( PLS) 、倾向评分法( Propensity score)、Lasso 回归、数据分组处理( group method of data handling，GMDH)
参考文献(并不全)：
Tolerance and condition in regression copmputations．
Diagnosing and dealing with multicollinearity．
Multicollinearity: a bayesian interpretation．
Principle Components Regression in Exploratory Statistical Research．
Multicollinearity in regression analysis: the problem revisited．
Collinearity and least squares regression．
Modeling data tables by principal component and PLS: class patterns and quantitative predictive relations．
The central role of the propensity score in observational studies for causal effects．
Collinearity diagnosis for a relative risk regression analysis: an application to assesment of diet cancer relationship in epidemiological studies．
Multicollinearity and imprecise estimation．
Ridge regression: biased estmiation for nonorthogonal problems．
Multicollinearity caused by specification errors．
The fallacy of differencing to reduce multicollinearity．
Propensity score methods for bias reduction in the comparison of a treatment to a nonrandomized control group．
觉得最好的办法应该是先进行自变量的选择，而不是一上来就回归。
另外这个问题似乎经济学更感兴趣，感觉医学反倒兴趣不是特别大。没有仔细研究过这个问题，隐约觉得lasso可能是最简单的。反正近来的趋势就是lasso这类方法什么都可以往上靠。
