This is the seventh of 10 chapters on "Spatial Modeling II" of the user manual for CrimeStat IV, a spatial statistics package that can analyze crime incident location data.
This chapter, "Discrete Choice Modeling," describes the discrete choice framework and the two well-known models that are part of it: the Multinomial Logit (MNL) and the Conditional Logit (CL). These techniques require a thorough knowledge of statistical analysis, especially regression modeling. A background in economics is beneficial, but not necessary. Analysts who want to use these techniques, particularly the Conditional Logit model, are advised to consult an expert in developing applications. The MNL and CL are closely related statistical regression models that can be used to analyze a discrete outcome variable as a function of a set of independent variables. Discrete variables are also known as nominal or categorical variables. They can take on a finite number of unordered, mutually exclusive values. Both the MNL and the CL are generalizations of the logit model, which is used to analyze binomial (two category) outcome variables. Gender is an example of a binomial variable (either male or female). Although the MNL and CL models can be used for all analytical problems when the outcome variable is discrete (nominal, categorical), in a number of disciplines, the models are used to study the way that people or organizations make choices. Many research questions in the social and behavioral sciences, including criminology, deal with understanding and predicting discrete choices. Choice is also a central concern in crime analysis. The CrimeStat discrete choice module is designed for regression when the dependent variable consists of unordered categories such as type of weapon or neighborhood where a crime is committed. 9 tables, 21 references, and attached paper on modeling correlates of weapon use in Houston (Texas) robberies with the Multinomial Logit model