Dummy variable - Main points Part I (Applied Econometrics)

 □ Dummy Variable

- Dummy Explanatory variables can be used in tests for the coefficient stability in the linear regression models, for obtaining predictions in the linear regression models, and for imposing cross equation constraints.

- One should exercise caution in using the dummy variables when there is heteroscedasticity or autocorrelation. In th presence of the autocorrelated errors, tge dummy variables for testing stability have to be defined suitably.

- Regarding the dummy dependent variables, there are three different models that one can use: the linear probability model, the logit model, and the probit model. The linear discriminant function are just proportional to those of the linear probability model. The coefficients of the discriminant function are just proportional to those of the linear probability model.

- For comparing the liner probability, logit and probit models, one can look at the number of cases correctly predicted. However, this is not enough. It is better to look at some measures of R ^ 2. The remedial measures of R ^ 2: squared correation between y and yhat, Effron's R ^ 2, Cragg and Uhler's R^2, and Mc Fadden R ^ 2. 

For the practical purposes, the first two are descriptive enough. The computation if different R ^ 2 can be illustrated.

To be contd...

Aditya Pokhrel

MBA, MA Economics, MPA