Prediction of Spatial Cumulative Distribution Functions Using Subsampling

Abstract

The spatial cumulative distribution function (SCDF) is a random function that provides a statistical summary of a random field over a spatial domain of interest. In this article we develop a spatial subsampling method for predicting an SCDF based on observations made on a hexagonal grid, similar to the one used in the Environmental Monitoring and Assessment Program of the U.S. Environmental Protection Agency. We show that under quite general conditions, the proposed subsampling method provides accurate data-based approximations to the sampling distributions of various functionals of the SCDF predictor. In particular, it produces estimators of different population characteristics, such as the quantiles and weighted mean integrated squared errors of the empirical predictor. As an illustration, we apply the subsampling method to construct large-sample prediction bands for the SCDF of an ecological index for foliage condition of red maple trees in the state of Maine.

Key Words:

• Environmental monitoring
• Increasing domain asymptotics
• Infill asymptotics
• Simultaneous prediction intervals
• Spatial subsampling