论文部分内容阅读
随着仪器仪表智能化程度的提高，在线性回归中寻求较好的粗差剔除算法变得越来越重要。通常的方法是在所有观测值的基础上作一初始的线性回归，然后将离差最大的点作为可能的粗差点。该方法常由于粗差值参与初始回归造成的不准确而发生判别错误。本文提出一种在线性回归中用特征值方法判别粗差点的方法，它不依赖于初始的回归离差，从而提高了判别的准确性。该方法应用于智能离子计中有关线性回归校正法及Ｇｒａｎ法数据处理软件中，取得满意的效果。
With the improvement of the instrumentation intelligent level, it is more and more important to seek a better algorithm of gross error rejection in linear regression. The usual method is to make an initial linear regression based on all observed values and then use the point with the largest deviation as a possible gross outlier. This method is often due to the gross error involved in the initial regression caused by inaccuracy and discrimination. In this paper, we propose a method of eigenvalue method to determine the rough points in linear regression, which does not depend on the initial regression and thus improves the accuracy of the discriminant. The method is applied to the intelligent ion meter in the linear regression correction method and Gran data processing software, and achieved satisfactory results.