Unconditional quantile regression

[Ana Vasconcelos]

Hello,

Can you please help me with the command to run an unconditional quantile regression?
Thank you in advance.

[FernandoRios]

Hi Ana,
There are three commands to estimate Unconditional quantile regressions. All of them being community contributed. This commands are
"rifreg" (see https://faculty.arts.ubc.ca/nfortin/datahead.html)
"xtrifreg" (to install ssc install xtrifreg)
"rifhdreg" (to install ssc install oaxaca_rif)
xtrifreg and rifhdreg can be used to estimate UQR with Fixed effects.

See the examples in the help files to check the syntax.

Best
Fernando

[Ana Vasconcelos]

Hi Fernando,
Thank you very much!
Best regards

[Henry Nguyen]

Dear Stata Users,
Can you suggest how can I check robustness for equal of coeficient after I run "rifhdreg". For the qreg, I can check the different for quantiles (for example quantile 10 and 20 by command "te [q10=q20]") but in "rifhdreg", i can not. Thank for your help,

[FernandoRios]

Hi Henry
You could do that using the companion to rifhdreg call rifsureg. Bellow an example

*individual RIF regressions
rifhdreg lnwage educ exper tenure female, rif(q(20))
rifhdreg lnwage educ exper tenure female, rif(q(80))
*Simulteanous RIF regressions
rifsureg lnwage educ exper tenure female, qs(20 80)
test [q20=q80]:educ
test [q20=q80]:exper
test [q20=q80]:tenure
* Better yet. Bootstrap simultaneous RIF regressions (similar to sqreg)
bootstrap:rifsureg lnwage educ exper tenure female, qs(20 80)
test [q20=q80]:educ
test [q20=q80]:exper
test [q20=q80]:tenure

Also see
Rios-Avila, F. 2019. Recentered influence functions in Stata: Methods for analyzing the determinants of poverty and inequality. Working paper 927, Levy Economics Institute.
http://www.levyinstitute.org/pubs/wp_927.pdf or see rif_paper for the latest version.
Or the help files for other relevant examples.

HTH
Fernando

[Henry Nguyen]

Dear @FernandoRios Thank you so much.
Best regards,

[Henry Nguyen]

Dear Fernando,
Thank for your very detailed reply! I have two questions for you, please.
(1) Can you help me check robustness for "xtrifreg" like above, please? I used your code for "xtrifreg" but it did not work.
(2) I try to plot "qregpd" results for many times but I can not get result. I see your code (plot the quantile coefficients after running "rifhdreg") in another topic, I hope you can help me for "qregpd"!
Thank you so much,
Best regards,

[FernandoRios]

Hi Henry,
Unfortunately, that is not possible with a one-step command. xtrifreg uses xtreg to account for fixed effects, whereas my own rifhdreg uses reghdfe. As far as I know, neither of these commands are supported by suest, and sureg does not admit for (many) fixed effects.

The options that I could suggest is:
1. if you do not have many individual effects (say 100?), just add it to the rifsureg using factor notation:

 rifsureg lnwage educ exper tenure female i.age, qs(20 80)

2. If you have MORE fixed effects, you could bootstrap the whole procedure, and use those results.

program srifhdreg, eclass
 rifhdreg lnwage educ exper tenure female, rif(q(20)) abs(age)
 matrix b1=e(b)
 rifhdreg lnwage educ exper tenure female, rif(q(80)) abs(age)
 matrix b2=e(b)
 matrix coleq b1=q20
 matrix coleq b2=q80
 matrix b=b1,b2
 ereturn post b
end 
bootstrap, cluster(age) seed(1): srifhdreg

I havent use qregpd, but Im guessing if you use the code you mention, and change rifhdreg to qregpd, you should be able to get the plot that you want.
HTH