变差函数稳健性研究

Study on Variogram Robustness

地质勘探过程中得到的样品数据往往偏离正态分布,这种情况下估算得到的变差函数值往往会偏离变差函数的实际值.针对这一问题,通过分析样品中成矿元素的分布特点,提出了采用稳健估计方法研究成矿的空间变差性,计算其变差函数.通过幂变换使样品服从正态分布,在此基础上采用稳健估计中的中位数法和三点均值法分别计算变差函数.采用交叉验证对变差函数模型最优性进行检验,结果表明采用稳健估计方法计算变差函数能够使估计值具有更高的精度.

The sample data acquired in the process of geological exploration often deviate from normal distribution. So, the calculating values from relevant variogram will deviate from their real values. Analysing the distribution characteristics of ore-forming elements in samples, a method of calculating variogram by robust estimation is proposed to study the spatial ore-forming variability. Samples are available to obey normal distribution through exponential transform, then the variogram is calculated separately by median and tri-mean methods of robust estimation. The model optimization of variogram is verified according to cross-validation, and the result shows that the estimated value can get higher accuracy by means of robust estimation to calculate variogram.