* Reproducing Table 7.1 in the WandB book

use moraz.dta 

* Provides definition of variables
describe

* Summarizes the data
summarize

* Tobit estimation of the hours equation
tobit hours nwifeinc educ exper expersq age kidslt6 kidsge6, ll(0)

* OLS using all of the data 
 
regress hours nwifeinc educ exper expersq age kidslt6 kidsge6 

* OLS with all of the zero hours deleted

drop if hours == 0

regress hours nwifeinc educ exper expersq age kidslt6 kidsge6

craggit hours nwifeinc educ exper expersq age kidslt6 kidsge6, ///
        second( hours nwifeinc educ exper expersq age kidslt6 kidsge6)
 
* Likelihood Ratio test: Tobit (restricted model) versus Cragg (Tobin II) model
* (unrestricted model)  Logl(restricted) = -3819.0946, 8 parameters
* Logl(unrestricted) = -3791.9498, 16 parameters
* Chi-square (8 df) = -2(Logl(restricted) - Logl(unrestricted)
* -2(-3819.0946 -(-3791.9498)) = 54.2896 (p = 0.0000)
* Conclusion: The Cragg model is preferred.