Book contents
- Frontmatter
- Contents
- Preface
- 1 Insurance data
- 2 Response distributions
- 3 Exponential family responses and estimation
- 4 Linear modeling
- 5 Generalized linear models
- 6 Models for count data
- 7 Categorical responses
- 8 Continuous responses
- 9 Correlated data
- 10 Extensions to the generalized linear model
- Appendix 1 Computer code and output
- Bibliography
- Index
7 - Categorical responses
Published online by Cambridge University Press: 04 June 2010
- Frontmatter
- Contents
- Preface
- 1 Insurance data
- 2 Response distributions
- 3 Exponential family responses and estimation
- 4 Linear modeling
- 5 Generalized linear models
- 6 Models for count data
- 7 Categorical responses
- 8 Continuous responses
- 9 Correlated data
- 10 Extensions to the generalized linear model
- Appendix 1 Computer code and output
- Bibliography
- Index
Summary
Categorical variables take on one of a discrete number of categories. For example, a person is either male or female, or a car is one of a number of models. Other examples are colors, states of wellbeing, employment status and so on.
The simplest example of a categorical variable is where the outcome is binary, meaning it can take on only one of two values such as claim or no claim on a policy, dead or alive. These outcomes are usually coded as 0 or 1, with the occurrence of the event of interest (claim, death) coded as 1, and nonoccurrence as 0. The terms “success” and “failure” are also often used for the occurrence and non-occurrence of the event.
Categorical variables fall into two distinct classes: those whose categories have a natural ordering or otherwise. For recording purposes, it is often convenient to code the categories numerically. For example 1 may denote Red, 2 Blue, and so on. In this case the numbers are purely “nominal” and for example 2 does not mean better than 1. In injury classification, however, 1 may denote minor injury, 2 more major up to 8 indicating catastrophic injury and 9 death. In this case the categories are ordered, although it is not necessarily the case that 4 is twice as bad as 2.
- Type
- Chapter
- Information
- Generalized Linear Models for Insurance Data , pp. 97 - 119Publisher: Cambridge University PressPrint publication year: 2008