An asymptotic theory for the nugget estimator in spatial models

作者：Kim Tae Yoon^*; Park Jeong Soo; Song Gyu Moon

来源：Journal of Nonparametric Statistics, 2010, 22(2): 181-195.

DOI：10.1080/10485250903193997

• nonparametric nugget estimator
• central limit theorem
• risk calculation
• rate of convergence
• nearly infill domain sampling

The nugget effect is an important parameter for spatial prediction. In this paper, we propose a nonparametric nugget estimator based on the classical semivariogram estimator and describe its large sample distributional properties. Our main results are a central limit theorem and a risk calculation for the estimator when observations are made from a nearly infill domain sampling. From our results, we note that the performance of the estimator depends on the sampling design as well as the choice of bandwidth. In particular, we show that the estimator suffers from strong dependency when d, the dimension of the underlying spatial process, is less than or equal to 2a, a parameter related to the degree of smoothness and dependence of the underlying process. When d > 2a, however, the estimator turns out to achieve an optimal rate with an optimal choice of h. We report on the results of simulations to empirically study the estimator.

• 出版日期2010