ABSTRACT
This chapter looks at two data sets from very different application areas. The first consists of data collected during a study of a psychiatric screening questionnaire, the General Health Questionnaire (GHQ), designed to help identify possible psychiatric ‘caseness’. The second data set arises from an investigation into the reasons for mortgage default. In the first data set, interest lies in assessing whether GHQ score is predictive of ‘caseness’ and whether the sex of a subject plays a role in this prediction. In the second, the main question is whether any of the four explanatory variables might be used to identify mortgage loans at risk of default. A suitable approach to modelling a binary response is to use logistic regression. Logistic regression is available in S-PLUS via the generalised linear model function or by using the Logistic Regression dialog. Both possibilities are used in the chapter.
