Spatial estimation from sparse data
Moreno, Robert Medellin
Master of Arts
Leading researchers in methods of Spatial Statistics advocate applying kriging methods to samples of 100 or more observations. However, in practice it is not uncommon to have as few as 20 observations from which to estimate a surface. The focus of this thesis is to compare various kriging models, given a small number of observations and develop a method for finding a lag at which to fit a variogram model. Specifically, the spherical, rational quadratic, and exponential variogram models are examined. A comparison among kriging methods and kernel methods is presented.