ABSTRACT
This chapter provides an introduction to the analysis of functions indexed by spatial locations. It presents the fundamental concepts of spatial statistics of scalar observations. The chapter then defines stationary and isotropic random fields of functions. Functional kriging and the estimation of the mean function of spatially indexed curves are considered. Implementation of these procedures in the R package geofd is presented. There are many excellent monographs and textbooks on spatial statistics. The monograph of Cressie (1993) provides a comprehensive account of the subject. The more recent monograph of Cressie and Wikle (2011) focuses on spatio-temporal data emphasizing Bayesian methods, which are also treated in detail by Banerjee et al. (2004). In spatial spatistics, the term kriging refers to linear prediction. Spatial point process data are concerned with locations at which events occur. Estimation of the mean function is needed to implement the kriging method. For illustration, one use the well-known Canadian weather data set available in the package fda.
