Abstract

Incorporating the spatial structure of data from agricultural field experiments into inference procedures has become an important topic in recent years. As part of a larger project to determine whether or not reliable predictions and estimates can be obtained for sample sizes often encountered in traditional field experimentation, this paper focuses on the small sample estimation of the parameters of the exponential semivariogram model. Simulation studies were conducted for both expanding and fixed domains. The results indicate large sample to sample variation in sample and fitted semivariograms, neither of which may be "close" to the true model. Distributions of individual parameter estimators are skewed and highly variable. Empirical coverage levels for large sample confidence intervals for the parameters are well below the nominal level and, contrary to what would be expected, decrease as the sample size increases. The results cast doubt on the success of incorporating spatial structure into traditional field data analyses.

Keywords

exponential semivariogram, simulation, small sample estimation, spatial data

Creative Commons License


This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Share

COinS
 
Apr 27th, 10:45 AM

A SIMULATION STUDY OF EXPONENTIAL SEMIV ARlO GRAM ESTIMATION

Incorporating the spatial structure of data from agricultural field experiments into inference procedures has become an important topic in recent years. As part of a larger project to determine whether or not reliable predictions and estimates can be obtained for sample sizes often encountered in traditional field experimentation, this paper focuses on the small sample estimation of the parameters of the exponential semivariogram model. Simulation studies were conducted for both expanding and fixed domains. The results indicate large sample to sample variation in sample and fitted semivariograms, neither of which may be "close" to the true model. Distributions of individual parameter estimators are skewed and highly variable. Empirical coverage levels for large sample confidence intervals for the parameters are well below the nominal level and, contrary to what would be expected, decrease as the sample size increases. The results cast doubt on the success of incorporating spatial structure into traditional field data analyses.