区域化变量非正态分布的稳健性

Stationarity for the non-normal distribution of regionalized variables

从区域化变量分布的稳健性角度,借鉴产品加工质量精度控制的理论,基于Johnson分布曲线族提出了一套非正态数据转换方法,并设计了转换流程.以SPSS和Surpac矿业软件为工具,通过实际案例分析了传统数据转换方法和Johnson转换方法的应用效果.实践证明,采用上述方法转化后的数据开展统计分析、变异函数模型的拟合及验证、克立格估计等工作,更容易满足稳健性要求,并能从本质上减小估计误差,提高估计精度.

Based on Johnson distribution curves and the theory of product process quality accuracy control, a method to transform non-normal distribution data was proposed from the view of regionalized variable distribution, and the transformation flow was also designed. With SPSS and Surpac mining software as tools, the applied efficiency of traditional data and Johnson transform methods was analyzed through several cases. It is proved that by using the transformed data of the proposed method, the statistical analysis, fitting and validation of the variation function model, and Kriging estimation could meet the need of stationary, the estimation error can be minimized, and the precision of prediction can be improved.