I have just run a double-limit Tobit model for cross-sectional data. I would like to know the robustness checks that should be carried out before inference is made from the Tobit model results?
generate its residuals and do the diagnostic tests manually, like durban watson value for autocorrelation, hetroskedasticity test.
the manual testing can be learn from Gujarati basic Econometrics book.
If you have cross-sectional data, you shouldn't look for autocorrelation. You should use robust standard errors because of heteroskedasticity, if you have a large dataset. You can also analyze if your coefficients vary too much or not whenever you remove one explanatory variable from the model. And you have too think if there is a problem of endogenity (that is, a variable in the error term of the model, correlated with an explanatory variable). Wooldridge is an excellent book to follow.
- Badges
- Science topic
- Similar topics
- Geoscience
- Atmospheric Sciences
- Atmosphere