* ------------
* Conditinal Quantile Regression
* ------------

sum earning, d

qreg earning, q(25)
qreg earning, q(50)
qreg earning, q(75)

sum earning if white==1, d

* (1) Qtile for whites
qreg earning if white==1, q(25)
qreg earning if white==1, q(50)
qreg earning if white==1, q(75)

sum earning if white==0, d /* Asian */

* (2) Qtile for Asian
qreg earning if white==0, q(25)
qreg earning if white==0, q(50)
qreg earning if white==0, q(75)

* (3) The coefficient of Asian = (2) - (1) 
qreg earning Asian, q(25)
qreg earning Asian, q(50)
qreg earning Asian, q(75)


* Qreg coefficients can be directly transformed by log and anti-log. 

qreg earning i.Asian##i.edu, q(10)
qreg lnearning i.Asian##i.edu, q(10)

qreg earning i.Asian##i.edu, q(90)
qreg lnearning i.Asian##i.edu, q(90)

* whilte the coefficients of OLS differ depending on whether log-transformed or not. 

reg earning i.Asian##i.edu
reg lnearning i.Asian##i.edu

* Thus, the predicted qtile points will be identical whether you use log-transformed dependent variable or not.