多元广义岭估计及K值选取的 Q(c)准则

当自变量间存在复共线性时,最小二乘估计就表现出不稳定并可能导致错误的结果.本文采用广义岭估计β(K)来估计多元线性模型的回归系数β=vec(B),通过岭参数K值的选取,可使广义岭估计的均方误差MSE小于最小二乘估计的MSE.指出了广义岭估计中根据MSE准则选取K值存在的主要缺陷,采用了一种选取K值的新准则Q(c),它包含MSE准则和最小二乘LS准则作为特例,从理论上证明和讨论了Q(c)准则的优良性,阐明了c值的统计含义,并给出了确定c值的方法.

Generalized Multivariate Ridge Regression Estimate and Criteria Q(c) for Choosing Matrix K

When multicollinearity is present in a set of the regression variables, the least square estimate of the regression coefficient tends to be unstable and it may lead to erroneous inference. In this paper, generalized ridge estimate fl,(K) of the regression coefficient =vec(B) is considered in multivariate linear regression model. The MSB of the above estimate is less than the MSB of the least square estimate by choosing the ridge parameter matrix K. Moreover, it is pointed out that the criterion MSB for choosing matrix K of generalized ridge estimate has several weaknesses. In order to overcome these weaknesses, a new family of criteria Q(c) is adpoted which includes the criterion MSB and criterion LS as its special case. The good properties of criteria 0(c) are proved and discussed from theoretical point of view. The statistical meaning of the scale c is explained and the methods of determining c are also given.