This paper is from Session 3F: Teaching nonparametric methods
which comes under Topic 3: Statistics education at the post-secondary level Full topic list
(Thursday 6th, 14:00-15:30)
Smoothing techniques in spatial statistics
AbstractThe variogram is one of the most important tools in the assessment of spatial variability of a spatial statistical model. Estimation and testing on this function is a crucial problem in random processes inference, with several applications in a broad spectrum of areas such as geostatistics, hydrology, atmospheric sciences, etc. We show in this work how a generalized family of variogram estimators can be built based on the classical ideas of smoothing techniques in nonparametric regression. Some examples will be given in order to compare the performance of Nadaraya-Watson and Local Linear estimators with the empirical variogram. The proper choice of the bandwidth for these methods will be discussed. Some applications to atmospheric and/or environmental data will also be provided. Finally, some extensions to the space-time setting will be considered. Special emphasis will be placed on teaching aspects in this paper.