ABSTRACT
When the dependent variable has a limited set of potential values, such as being a dummy variable, then other types of models might be more appropriate and provide a better fit of the data. This chapter introduces these other methods, including: probits, logits, multinomial logits, and survival models. For a few of these, there is a discussion on the pros and cons of using the given method versus just using the simpler and more straightforward Ordinary Least Squares method. The chapter emphasizes that the choice of method does not matter if there are meaningful biases on the coefficient estimates (e.g., reverse causality) that were not addressed.
