主成份估计的特征根因子筛选方法

线性回归模型的主成份估计具有模型紧致和减弱病态的作用,并提高参数估计效果,但是如何筛选特征根却是主成份估计的重要问题。本文找到了主成份估计的特征根与残差平方和之间的确定关系,并根据模型拟合优度和减弱测量随机噪声对参数估计的影响,提出主成份估计筛选特征根的准则。由于特征根与残差平方和具有确定关系,使用这个准则就极为方便,它既可以选到最优组合,又大大减少了计算量,这就具有实际使用价值。

SIEVING METHOD FOR CHARACTERISTIC ROOT FACTOR OF PRINCIPAL COMPONENT ESTIMATE

The Principa Clomponent Estimate of the linear regression has the function of model compactnees and Weakening of abnormal state. It may also raise the efficiency in estimating paramaters. However, how to sieve characteristic roots is an important problem of the Principal Component Estimate. This paper has found the certain relation between the cha racteristic roots and residual sum of squares of Principal Component Estimate, and has advanced a criterion for sieving characteristic roots of Principal Component Estimate according to the fitting merits of the model and weakening the effection of measuring random noise to the estimate of parameters. Because characteristic roots and residue sum of squares have the certain relation, using these criterions is very convenient. Thus it not only may select the optimal combination, but also minimize the complexity of computation, so that it is valuable in practice.