Fixed effects - unconditional quantile regression

[Ana Vasconcelos]

Hello,

Can you help me with the following:
I am using the command "xtrifreg" with country fixed effects in the following way: xtrifreg Y X1 X2 X3, fe i(country_) q(25) clusterbootstrap reps (50).
How can I add the time fixed effects (variable name:qdate)?
Thank you in advance.
Best regards

[Liu Qiang]

Dear Ana, the newly proposed command -rifhdreg- may help you. You can get this command using

ssc install rif, replace

[Shrek Deng]

Dear Ana, the newly proposed command -rifhdreg- may help you. You can get this command using

ssc install rif, replace

Hello, do you know any way to plot the quantile coefficients after running rifhdreg?

[FernandoRios]

Hi Shrek
There is no command to do that, but you can probably adapt the following code to make the figures you are looking for:

clear all
use http://fmwww.bc.edu/RePEc/bocode/o/oaxaca.dta, clear
 

capture matrix drop rbs rll rul
 forvalues i=10(5)95 {
 qui:rifhdreg lnwage educ exper tenure age , rif(q(`i')) robust
 matrix rtb=r(table)
 matrix rbs=nullmat(rbs)\[`i',rtb["b",.]]
 matrix rll=nullmat(rll)\[`i',rtb["ll",.]]
 matrix rul=nullmat(rul)\[`i',rtb["ul",.]]
}
 
 svmat rbs
 svmat rll
 svmat rul
 local m: colfullnames rbs
 foreach i of local m {
 local ii=`ii'+1
 label var rbs`ii' "`i'"
 label var rll`ii' "`i'"
 label var rul`ii' "`i'"
 }
 
 label var rbs1 quantile

 two rarea rll3 rul3 rbs1 || scatter rbs3 rbs1 , connect(l)
 two rarea rll2 rul2 rbs1 || scatter rbs2 rbs1 , connect(l)
 two rarea rll5 rul5 rbs1 || scatter rbs5 rbs1 , connect(l)

HTH
Fernando

[Henry Nguyen]

Hi @
FernandoRios

Do you know any way to plot the quantile coefficients and xtreg (fe) after running rifhdreg and xtreg in one graph. I mean that the graph shows both results of xtreg (fixed effect) and rifhdreg.
Thank you so much,

[FernandoRios]

Only the code I posted before
i have written a code to do that automatically

[Henry Nguyen]

Hi FernandoRios,
Thank you for your response.
Best regards,

[Shrek Deng]

Hi Shrek
There is no command to do that, but you can probably adapt the following code to make the figures you are looking for:

clear all
use http://fmwww.bc.edu/RePEc/bocode/o/oaxaca.dta, clear


capture matrix drop rbs rll rul
forvalues i=10(5)95 {
qui:rifhdreg lnwage educ exper tenure age , rif(q(`i')) robust
matrix rtb=r(table)
matrix rbs=nullmat(rbs)\[`i',rtb["b",.]]
matrix rll=nullmat(rll)\[`i',rtb["ll",.]]
matrix rul=nullmat(rul)\[`i',rtb["ul",.]]
}

svmat rbs
svmat rll
svmat rul
local m: colfullnames rbs
foreach i of local m {
local ii=`ii'+1
label var rbs`ii' "`i'"
label var rll`ii' "`i'"
label var rul`ii' "`i'"
}

label var rbs1 quantile

two rarea rll3 rul3 rbs1 || scatter rbs3 rbs1 , connect(l)
two rarea rll2 rul2 rbs1 || scatter rbs2 rbs1 , connect(l)
two rarea rll5 rul5 rbs1 || scatter rbs5 rbs1 , connect(l)

HTH
Fernando

Thank you, Fernando. Sometimes, when I run regressions with rifhdreg, I encountered with r(2001): insufficient observations. Do you have any idea what could be the problems?

[FernandoRios]

The most likely scenario is that you are using an incorrect index for the quantile.
For example, if you type

rifhdreg y x, rif(q(0.10))

It will not estimate the 10th quantile, but the 0.1th quantile.
The correct code would be:

rifhdreg y x, rif(q(10))

Also, if you are trying to make plots, perhaps you may want to look into -qregplot- (ssc install qregplot).
Fernando

[Shrek Deng]

Thanks for your suggestion for plotting. But I enter the index for quantile correctly. Could it be that I control too many fixed effects? As I reduce the number of variables absorbed, the error disappeared. But it works for reghdfe even when I control lots of fixed effects. So confusing.

[Adrian Santonja]

Hi Shrek,

I had the same problem recently. The problem seems to occur when you include string variables in the abs() option of rifhdreg. When I encoded those string variables, I was able to run rifhdreg without error. You are right, however, that this is not the case under the current version of reghdfe.

Hope it helps!
Adrian