Dependent Dummy Variable Models: An Application of Logit, Probit and Tobit Models on Survey Data

Abstract

In the current study, logit, probit and tobit models which are commonly used among dependent dummy variable models are included. These models are also known as limited dependent variable models in the literature. Surveys, which are widely used in the field of social sciences, are carried out with limited options due to their nature. Linear regression models cannot be used in statistical estimations since they do not provide assumptions in limited analysis. In this case, different regression models may be preferred. The main purpose of this study is to compare the Tobit model used in censored data and the binary logit and binary probit regression models derived from this model. For this purpose, analysis were conducted on survey data. Logit, probit and Tobit model coefficient estimates and marginal effects were calculated. In addition, AIC and BIC values were obtained from the model selection criteria for these 3 models.