ABSTRACT
In the analysis of nonspatial data sets the methods discussed in this chapter are not necessarily considered exploratory. We classify them as exploratory here because in parameter estimation and signi–cance testing, spatial data may violate the assumptions underlying the traditional regression model, and, therefore, the estimate or signi–cance test may not be valid. Our initial use of regression will be to gain further insights into the relationships among the data. Con–rmatory analysis will be carried out in later chapters in which the analysis incorporates assumptions about the spatial relationships among the data. Section 9.2 provides an introduction to multiple linear regression and develops the approach to model selection that we will use. Section 9.3 applies this approach to the construction of alternative multiple linear regression models for yield in Field 1 of Data Set 4. Section 9.4 introduces generalized linear models (GLMs) and shows how these are applied to data from Data Sets 1 and 2.
