化学计量学中的稳健估计方法

稳健统计学是八十年未才基本定型的统计学分支，它是针对实际情况中假设模型常常只是对实际数据的一种近似而导致传统统计学推断失误而发展起来的。稳健统计学构造一些新的具有稳健性的方法，使得在假设模型满足时，稳健方法具有接近最优的性能；在实际数据与假设模型有差别时，其性能仍为次优的；而在实际数据与假设模型差别大时，统计方法的性能也不会变得过差。本文介绍了稳健估计的一般概念。综述了化学计量学中的稳健估计方法，

Robust Estimation Methods in Chemometrics

Robust statistics as a new branch of chemometrics was developed to deal with the discrepancy between the assumed model and the data structure for which conventional statistics failed. The performance of a robust chemometrics method should be optimum or quasi -optimum when there is no departure between the assumed model and real data, it should be impaired only slightly for small deviation from the assumption;and the deterioration of the method's performance should not be catatrophic even when the data depart substantially from the assumed model. In this paper, the basic concepts of robust statistics are intro-duced, and the robust estimation methods developed and applied in chemometrics are re-viewed with 53 references. The study of robust methods in the treatment of multivariate data has also been highlighted.