/* Examples of dummy variables 
   Data " cps_msa2013 : 2013 Current population survey for 3 MSA
   3 MSA : NY, LA, CHICAGO
   
   Created by Jin Man Lee
   Updated: 2/17/2015
 */  

use http://bigblue.depaul.edu/jlee141/econdata/cps_data/cps_msa2013

gen lhrwage = log(hrwage) 
label var lhrwage "Log of Nominal Hourly Wage"
/* 
1. Creating continuous education variable
   Number of Years of Education 
*/
// tablulate the educ92 
tab educ92
// tablulate the educ92 without label 
tab educ92, nolabel 
 
gen educyr = 0 
replace educyr =  1 if educ92 == 1 
replace educyr =  4 if educ92 == 2 
replace educyr =  6 if educ92 == 3 
replace educyr =  8 if educ92 == 4 
replace educyr =  9 if educ92 == 5 
replace educyr =  10 if educ92 == 6 
replace educyr =  11 if educ92 == 7 
replace educyr =  12 if educ92 == 8  |  educ92 == 9  
replace educyr =  14 if educ92 == 10 | educ92 == 11 | educ92 == 12 
replace educyr =  16 if educ92 == 13 
replace educyr =  18 if educ92 == 14  
replace educyr =  20 if educ92 == 15 
replace educyr =  21 if educ92 == 16 
 
/* 
2. Creating simple dummy variable 
  Here are some examples to create simple dummy variables 
*/

gen young = 0 
replace young = 1 if age < 25 

gen retired_age = age > 65 
gen retired_man = (age > 65 & female == 0 )

/* 
3. Creating multiple dummy variables  
  Here are some examples to create multiple dummy variables 
  wbho (white, black, hispanic, others) 
    -> generate four dummy variables (race1, race2, race3, race4) 
*/

tab wbho 
tab wbho, gen(race)
sum race1-race4

/* 
4. Regression using dummy variables aid factor variables
 */

regr lhrwage educyr female
regr lhrwage educyr race2-race4 
test race2 race3 race4

regr lhrwage educyr i.wbho  
predict fit1 
tw (line fit1 educyr if race1 == 1 ) ///
   (line fit1 educyr if race2 == 1 ) ///
   (line fit1 educyr if race3 == 1 ) ///
   (line fit1 educyr if race4 == 1 ), ///
   legend(lab(1 "Fitted for White") lab(2 "Fitted for Black") ///
          lab(3 "Fitted for Hispanic") lab(4 "Fitted for Others")) ///
   ytitle("Fitted log of hourly wage")  ///
   xtitle("Number of years of education")

gen female_educyr =  female*educyr 
regr lhrwage female  educyr female_educyr
regr lhrwage i.female##c.educyr
predict fit2

tw (line fit2 educyr if female == 0)   ///
   (line fit2 educyr if female == 1),  ///
   legend(lab(1 "Fitted for Male") lab(2 "Fitted for Female")) ///
   ytitle("Fitted log of hourly wage")  ///
   xtitle("Number of years of education")