几种特异值处理方法的比较

COMPARISONS AMONG THE METHODS OF HANDLING OUTLIERS

特异值（又称特高品位）存在于抽样调查之中。在地质统计学中，如果观测值存在有特异值，就严重的影响变差函数的计算结果，从而大大影响了地质统计学研究结果的精度。本文通过对目前国内外处理特异值方法（１．估计邻域法ＥＮＭ２．影响系数法ＩＣＭ３．相对变差函数法ＧＲＶ．ＰＲＶ）的比较，确定处理方法的优劣，对提高地质统计学研究结果的精度。

Outliers ( also called as extreme high grade data) may be present in sampling. In geostatistics, if sample values contain outliers, they will influence the result of variogram and the precision of forecast of geostatistics. This paper shows comparisons in handling outliers (1. Estimation Neighbourhood Method. 2. Influence Coefficient Method. 3. Relative Variogram Method) and gives a better method to handle outliers. It is very useful to improve the precision of forecast of geostatistics.