set matsize 10000

cd "C:\Users\sovid\OneDrive - The University of Kansas\Lectures\Soc910_AdvStat\examples"
use soc910_data_3_KLI_FEM.dta, clear

egen count = count(pid), by(pid)
keep if count>=2

xtset pid year

xtsum lnwage age female

reg   lnwage age female i.educ
xtreg lnwage age female i.educ, fe
xtreg lnwage age female i.educ, fe vce(r)
 
reg   lnwage age female i.educ       if pid<20000
est store OLS
xtreg lnwage age female i.educ       if pid<20000, fe vce(r)
est store m1
reg   lnwage age female i.educ i.pid if pid<20000, vce(r)
est store m2

estout OLS m1 m2

xtreg lnwage c.age##i.female i.educ, fe vce(r)



* Group-specific mean-centering for independent variables 
* You can estimate time-invariante variables along with time-variant variables. 

egen mlnwage = mean(lnwage), by(pid)
gen  clnwage = lnwage - mlnwage

egen mage    = mean(age), by(pid)

tab educ, gen(educ)
egen meduc1  = mean(educ1), by(pid)
egen meduc2  = mean(educ2), by(pid)
egen meduc3  = mean(educ3), by(pid)
egen meduc4  = mean(educ4), by(pid)

gen cage = age-mage
gen ceduc1 = educ1 - meduc1
gen ceduc2 = educ2 - meduc2
gen ceduc3 = educ3 - meduc3
gen ceduc4 = educ4 - meduc4

reg lnwage cage ceduc2-ceduc4 female mage meduc1-meduc4 

reg lnwage c.cage##i.female ceduc2-ceduc4 


* Hausman test
xtreg lnwage age female i.educ, fe v(r)
estimates store fe
xtreg lnwage age female i.educ, re v(r)
estimates store re
hausman fe re



reg   lnwage age female i.educ
estimate store OLS
xtreg lnwage age female i.educ, fe vce(r)
estimate store FEM
xtreg lnwage age female i.educ, re vce(r)
estimate store REM
xtreg lnwage age female i.educ, be
estimate store BEM

estout OLS FEM REM BEM, cells(b(star fmt(%9.4f)) se(par))      ///
        stats(r2_a N, fmt(%9.4f %9.0g) labels(R-squared))      ///
        legend label collabels(none) varlabels(_cons Constant)