摘要
为了改进存在复共线性的回归模型中回归系数的最小二乘估计的不足,利用构造岭估计的思想,只修正非常接近于零的那部分特征值,从而给出了回归系数的部分岭估计.在均方误差意义下,存在岭参数,使得回归系数的部分岭估计优于最小二乘估计.部分岭估计只修正了很少一部分特征值,所以相对于普通岭估计和广义岭估计的岭参数的确定,部分岭估计的岭参数的确定更加方便精确.
To improve the least squares estimate of regression coefficients in linear regression model which has the problem of the multicollinearity.We take advantage of the methods used in the ordinary ridge estimate,and give the partial ridge estimate of the regression coefficients.In the sense of mean squared errors,there exist ridge parameter,make the partial ridge estimate better than the least squares estimate.Because we only change the eigenvalues which are very close to zero in the partial ridge estimate,the determinant of the ridge parameter in the partial ridge estimate is easier than that in the ordinary ridge estimate and in the generalized ridge estimate and more precise.
出处
《河南理工大学学报（自然科学版）》
CAS
2011年第6期749-752,共4页
Journal of Henan Polytechnic University（Natural Science）
关键词
线性回归模型
复共线性
最小二乘估计
部分岭估计
均方误差
linear regression model
multicolinearity
least squares estimate
partial ridge estimate
mean squared errors
