克利金空間推估與複迴歸之關係及半變異元函數特性之探討

克利金推估為空間推估之重要方法之一，其中包括線性推估如簡單克利金推估、一般克利金推估、指標克利金推估、通用克利金推估等多種不同形式，及非線性之非聯結克利金推估等。就線性推估方式而言，簡單克利金及一般克利金對參數統計特性之需求與複線性迴歸有近似之處，但對隨機變數本身之統計特性則明顯不同。本文針對簡單克利金、一般克利金及複線性迴歸之異同從理論探討之，以說明應用這些方法之適當情況。另外亦對克利金推估中變異元函數的特性所代表之意義詳加闡述，以有助於半變異元模式及參數之選用。

關鍵字

克利金推估；複迴歸；半變異元

並列摘要

Kriging estimation is one of the most important methods of spatial estimation. The family of kriging estimators includes linear estimator like simple kriging, ordinary kriging, indicator kriging and universal kriging, as well as nonlinear estimator of disjunctive kriging. As a linear estimator, simple kriging and ordinary kriging require similar statistical properties for their parameters as does the multiple linear regression. However, the statistical properties of random variables are different for kriging and multiple linear regression. This paper discusses the similarity and assimilarity of kriging and multiple linear regression and help to clarify situations in which these methods are to be adopted. In addition, the implication of variogram behavior is interpreted and will benefit the selection of variogram model and parameters.

並列關鍵字

Kriging ； Multiple regression ； Variogram

主題瀏覽