User’s Guide : Advanced Single Equation Analysis : Discrete and Limited Dependent Variable Models
Discrete and Limited Dependent Variable Models
The regression methods described in “Basic Regression Analysis” require that the dependent variable be observed on a continuous and unrestricted scale. It is quite common, however, for this condition to be violated, resulting in a non-continuous, or a limited dependent variable. We will distinguish between three types of these variables:
qualitative (observed on a discrete or ordinal scale)
censored or truncated
integer valued
In this chapter, we discuss estimation methods for several qualitative and limited dependent variable models. EViews provides estimation routines for binary or ordered (probit, logit, gompit), censored or truncated (tobit, etc.), and integer valued (count data) models.
EViews offers related tools for estimation of a number of these models under the GLM framework (see “Generalized Linear Models”). In some cases, the GLM tools are more general than those provided here; in other cases, they are more restrictive.
Standard introductory discussion for the models presented in this chapter may be found in Greene (2008), Johnston and DiNardo (1997), and Maddala (1983). Wooldridge (1997) provides an excellent reference for quasi-likelihood methods and count models.