The course is really split between two components: maximum likelihood estimation and a variety of models to which it can be applied. Many of these models fall into a larger class known as generalized linear models. Among the models considered will be binary dependent variables, ordered and unordered discrete choice models, duration models, and selection models. While the title of the course is maximum likelihood, it is at least as much about these applications.
These models have wide-ranging applications in political science and many other ﬁelds. They also have something else in common: they cannot be estimated correctly using OLS, but can be with maximum likelihood. One of the great advantages of maximum likelihood is that it provides a uniﬁed framework for estimating a huge variety of models.
These models are of great importance to work in political science and other social sciences. While OLS is an incredibly useful tool, it will not be appropriate for much—if not most—of the data you encounter. For example, many of the dependent variables we wish to use involve discrete choices (e.g., voting decisions), counts (e.g., number of coups per year), durations (e.g., length of time between conﬂicts), and so on. The models discussed in this class can handle these situations and more. In addition to studying maximum likelihood and related models, we will look at examples of its applications within political science.