Generalized Additive Models,Graphical Diagnostics,and Logistic Regression

Generalized additive models (GAMs), although little known in geographical analysis, have considerable utility. In particular, they allow the conventional linear relationships of multiple regression to be generalized to permit a much broader class of nonlinear, but still additive, relationships between response and predictor variables. This paper illustrates how GAMs can be extended to the important case of logistic regression with binary response. Details of the distinctive form of logistic regression GAMs are provided, and appropriate means of estimation are outlined. The paper provides the first use, in the context of a logistic regression GAM, of the expansion method of assessing parameter variation. These techniques are illustrated by a reanalysis of information on hydrocarbon exploration wells drilled in south-central Kansas.