杠杆点存在时的稳健多元校正 ,是分析化学中有意义的课题。本文基于最小中位方差估计法机理 ,通过定义主灵敏度矢量来删除奇异点 ,进行迭代运算 ,获得一初始估计 ,检验后获得最终解答。本方法对仿真实验及实际多组分药物的分析均取得了满意结果 ,有效地解决了回归问题中强杠杆点的掩蔽效应 ,表明该方法是QSAR和多元校正研究中一个很有前景的稳健估计方法 Robust multivariate calibration in the presence of leverage points is a significant topic in analytical chemistry. Based on the principle of minimum mean square error estimation, we delete the singular points by defining the main sensitivity vector and perform iterative operations to obtain an initial estimate. After the test, the final solution is obtained. The proposed method achieves satisfactory results for the simulation experiments and the actual analysis of the multi-component drugs, and effectively solves the masking effect of the strong leverage points in the regression problem. It shows that this method is a promising robust estimation in the QSAR and multivariate calibration studies method