ABSTRACT
The previous chapter considered estimation of the parameters of a geostatistical model by a combination of method-of-moments and least squares methods. Those methods, collectively known as “classical geostatistics,” are relatively simple and do not explicitly require any distributional assumptions, but they are not optimal in any known sense. In this chapter, we present the estimation of parametric geostatistical models by likelihood-based approaches, which adhere to the likelihood principle (of course!) and, under appropriate regularity conditions, may be expected to have certain optimality properties.
