17 Limited Dependent Variable Models

There are slides associated with this chapter:

This chapter covers three regression models in which the dependent variable is somehow limited:

  1. Truncation: With truncated data, the researcher does not observe values past a particular point and those values are also not reported. Examples of truncation are low-income household studies, on-site visitation data, or time-of-use pricing experiments (excludes low-usage households).

  2. Censoring: In the case of censoring, values that are above or below a certain value are replaced by that value. For example, the demand for a particular class is not fully observed (absence of a waiting list). This is also called a Tobit model.

  3. Count dependent variable

The following packages are necessary for this section: AER truncreg, censReg, and pscl.

Truncation and censoring lead to a bias in the estimates. It is not always clear why or if the data is limited in its range.