线性回归中粗差点判别法的改进及其应用

Amelioration of Outlier Detection in Linear Regression and Its Application

线性回归是智能仪器数据处理中常用的方法,但可能存在的粗差点将降低其结果的准确性。因此,在线性回归中寻求较好的粗差剔除算法变得越来越重要。本文提出一种判别线性回归可疑粗差点的改进方法,从而有效地提高了判别结果的准确性。该法应用于智能离子计标准曲线法及Gran法等数据处理的软件设计,取得了满意的结果。

The rejection of outlier in linear regression is important in the data processing of intelligent instruments. A better method on outlier detection in linear regressions is described. This method avoids wrong detection of outlier in conventional methods, that is based on residual estimate for determining suspect observation in linear regression. The method is used in the software design of intelligent ion-analyzer for the data processing of standard curve method and Gran's method with satisfactory results.