摘要
针对最小二乘回归方法,回归系数的显著性F检验,如果F>临界值,则线性回归显著,但当多元回归模型的设计矩阵存在近似线性相关(称为复共线),即—XTX|≈0.这时这一判别标准不具备普遍意义。本文结合实例对设计矩阵出现复共线关系进行岭回归。
F (verification) is a important method to scale markedness of least-squares. If F〉boundary value, regression will be obvious. But when there is near-linear dependency(multicollinearity) in design matrix of multi-variate regression, |X^TX|≈0,This discrimination is not available. Writer want to linear regress matrix of near-linear dependency by ridge regression.
出处
《武汉科技学院学报》
2005年第10期51-53,共3页
Journal of Wuhan Institute of Science and Technology
关键词
最小二乘法
回归分析
回归系教
复共线
岭回归
least-squares procedure
regression analysis
regression coefficient
multicollinearity
Ridge regression