Hello.. FYI for those of you who have tried to use xtqreg , but had issues running it with an invalid syntax response - I contacted the author recently. He found a problem with one of the versions, which he has proceeded to fix. So reinstall the package with this command

net install xtqreg, from("http://www.imm.ki.se/biostatistics/stata") force replace

and ensure that the Rterm_path is correctly specified (you can use the sample dataset and codes provided earlier in this discussion, or obtain it in help xtqreg). Hope this helps!

addit: FYI - if you are modelling categorical variables, there is a slight variation to the syntax because the codes need to "call" on R. These are the codes that the author shared with me. I've got his permission to share this - hopefully this will help those who are stuck like I was :).

global Rterm_path "C:\Program Files\R\R-3.3.1\bin\x64\Rterm.exe"

use http://www.imm.ki.se/biostatistics/data/wtloss, clear

reshape long y, i(id) j(month)

xtset id month

xi, prefix(I): xtqreg y i.prog*i.month

If you have longitudinal data, then -qreg- is not the appropriate model (in the same way that -regress- is not the appropriate model for longitudinal data). Neither approach accounts for the autocorrelated nature of the data.

There appears to be a user-written program called -xtqreg- but I can't find it (other than in the slides below). I am going to contact the author and ask for a copy of the command. I'll let you know.

you may also want to consider another user-written program for Stata called -qrpd- tha performs quantile regression for panel data. I have not yet tried this program out.

http://works.bepress.com/david_powell/1/

Thanks Ariel. I am trying the qrpd option. It runs the model but cannot show the output when I use e(beta1) or e(se). It would be great help if you find the xtqreg.

I found it:

net install xtqreg, from(http://www.imm.ki.se/biostatistics/stata) replace

 You can check the help file (help xtqreg), after you have downloaded it.

In its current version the command simply calls the “lqmm” library in R.

Therefore, you need to have R with the “lqmm” library installed on your computer.

Please be reminded that the command is not released for public use and may contain bugs.

I have finally got the fixed effect quantile regression output using rqpd in R. But, this package does not perform the Machado-Santos Silva test for heteroskedasticity immediately after the regression fit (as it is done in Stata). Could anyone help how this test can be performed in R after fitting rqpd? Thanks for any help.

Sorry, I only know Stata...

Hello, Ariel. i am also looking for quantile regression with panel data by STATA.

The link  http://www.imm.ki.se/biostatistics/stata seems unavailable, could you provide other ways to have the xtqreg?

Qiao,

The link works fine. You need to type the following, from the command line in Stata:

net install xtqreg, from(http://www.imm.ki.se/biostatistics/stata) replace

I hope this helps

Ariel

Ariel,

Yes, I made a stupid mistakes, thanks so much. I have installed it.

however, when I run xtqreg, it shows 'the R script did not run, check the R path'? I do not understand what it means. Can you help to figure it out?

In addition, my laptop has R installed already, what should I do as well? or is there requirement for the version of R?

Qiao,

I don't know R, so I can't provide any more info. I suggest that you contact the authors of the command and ask them to trouble-shoot for you.

Ariel

Ariel,

Thanks all the same.

I will try to contact the authors. Hope they can help to figure it out.

Hi Qiaowen,

I also find the same message: 'the R script did not run, check the R path' when I tried xtqreg in STATA. Could you solve it? What is the issue?

here is the do file example:

clear

set more off

global Rterm_path "C:\Program Files\R\R-3.2.2\bin\i386\Rterm.exe"

//make sure you check you R version.. my is R-3.2.2 and intall lqmm package in

//your R program

use http://www.imm.ki.se/biostatistics/data/wtloss, clear

reshape long y, i(id) j(month)

gen inter = month*prog

xtset id month

// Random intercept

xtqreg y prog month inter

// Random intercept and multiple quantiles

xtqreg y prog month inter, q(25 50 75)

// Random intercept and random slope

xtqreg y prog month inter, random(month)

Thanks Vilaphonh for sharing this. I'll try this code.

Hi Qiaowen,

Have you solved this issue? I put lqmm in the library of R but still cannot run the R script and get results.

Thanks !

Hi Zilong Zheng,

Did you try the code I posed above. As you didn't give a detail of your problem, I don't know what it is and I am happy to check your code if you like..

Attached article might help.

Thanks Basheer, I tried this Rand article, it did not work for me, I am looking for coding that used two step method for fixed effect quantile regression proposed by Ivan Canay (2011). This method first perform a linear fixed effect estimation, then does a quantile regression. If anybody had any clue of this coding, please share. 

After trying xtqreg in STATA,it shows"invalid syntax".But I do install "xtqreg"and R with"lqmm".How to solve it??

 Hi Ziwei, you may like to try rqpd package with R downloadable from Prof Koenkar's website http://rqpd.r-forge.r-project.org/, this package works but the estimator has some limitations. On the other hand you may like to use Canay's two step estimator downloadable from his research page http://sites.northwestern.edu/iac879/research/

Dear Ujjwal,thank you soooo much and I will try it.

qregpd – quantile regression with panel data in Stata

qregpd implements the quantile estimator for panel data developed by Powell (2015). As detailed in Powell (2016) – an awesome paper - this estimator is a special case of the generalized quantile estimator implemented by genqreg. The estimator addresses a fundamental problem posed by alternative fixed-effect quantile estimators: inclusion of fixed effects alters the interpretation of the estimated coefficient on the treatment variable.

Examples:

Robust quantile regression for panel data:

Code:

. webuse nlswork

. qregpd ln_wage tenure union, id(idcode) fix(year)

Same as above, MCMC:

Code:

. qregpd ln_wage tenure union, id(idcode) fix(year) optimize(mcmc) noisy draws(1000) burn(100) arate(.5)

Robust instrumental variable quantile regression for panel data, grid-search optimization:

Code:

. qregpd ln_wage tenure union, id(idcode) fix(year) optimize(grid) min1(0) max1(0.06) intvl1(0.005) min2(0.05) max2(0.1) intvl2(0.005) instruments(ttl_exp wks_work union)

Further examples and options can be found in the help file. qregpd can be install via ssc (ssc install qregpd).

Enjoy! And of course, comments, criticisms, and suggestions are welcome...

hi Aymen,

thanks for sharing the new package. but I fail to install it. can you check out the screen shots below for me? 

thanks very much.

Sorry about that - there is in fact a dependency that I did not mention: you need to have the moremata package installed. I believe you can just get this up and running by typing in the command ssc install moremata.

Hope that helps!

Dear Aymen, thanks for sharing the codes. However when i am executing the stata codes, it showing that ": 3499 mm_panels() not found". Could you please suggest me how i can solve this problem. 

Many thanks!

@ Debasish Kumar Das 

It looks as though you do not have the amcmc package installed, which is required for any of the Markov-chain-Monte Carlo estimations, as mentioned in the dependencies.

Try doing:

Code:

ssc install amcmc

and then run your command and see if that works!

Best,

A.A

Thanks a lot Aymen! its working now!

could you please help me which code i could run with different quantiles ( 0.25, 0.50, 0.75) in MCMC method of panel quantile.

many thanks!

@ Aymen Ammari

I work on Powell panel quantile estimation.  I see that authors use noisy draws(1000) burn(100) arate(0.5). Is it so strict? In which criterion is it based on? Does the size of sample matters? In addition, do you know maybe how to calculate rmse for this model ?

Many thanks

@ Periklis Boumparis

well the author used 

1- Robust quantile regression for panel data: Standard using (qregpd )

2- Robust quantile regression for panel data with MCMC: using (adaptive Markov chain Monte Carlo)

MCMC is a way to solve high dimensional integral calculus by creating a map of the data. Imagine you have a mountain range and you need to know its dimensions. MCMC can do this by walking over the mountains and keeping track of the shape. It means it takes a very long time but you'll eventually get very accurate results

how to calculate rmse:

1- Calculate the difference between the observed and predicted dependent variables

2- Square them

3- Add them up, this will give you the "Error sum of squares," 

4- Divide it by the error's degrees of freedom, this will give you the "Mean error sum of squares," 

5- Take a square root of it, and this is the Root MSE

6- Done

Hope this help

@Aymen Ammari

Thanks for your help.

Do you calculate rmse for your model? If no, how do you assess your model? I am not sure if rmse is an appropriate measure for quantile models. Probably, we can use it only for 0.50 quantile.  When I calculate rmse, I get very high values, epecially for high and low quantiles.

Hello.. FYI for those of you who have tried to use xtqreg , but had issues running it with an invalid syntax response - I contacted the author recently. He found a problem with one of the versions, which he has proceeded to fix. So reinstall the package with this command

net install xtqreg, from("http://www.imm.ki.se/biostatistics/stata") force replace

and ensure that the Rterm_path is correctly specified (you can use the sample dataset and codes provided earlier in this discussion, or obtain it in help xtqreg). Hope this helps!

addit: FYI - if you are modelling categorical variables, there is a slight variation to the syntax because the codes need to "call" on R. These are the codes that the author shared with me. I've got his permission to share this - hopefully this will help those who are stuck like I was :).

global Rterm_path "C:\Program Files\R\R-3.3.1\bin\x64\Rterm.exe"

use http://www.imm.ki.se/biostatistics/data/wtloss, clear

reshape long y, i(id) j(month)

xtset id month

xi, prefix(I): xtqreg y i.prog*i.month

Hello,does anybody run the quantile regression with qregpd in Stata?I found when I run more times with the same quantile,the outcome in every time is different.Is there any problem with my performance?

Dear @Debasish Kumar Das 

Did you figure out the solution to the error "": 3499 mm_panels() not found"? Could you tell me how to fix it?

Thanks a lot.

can anyone send me a detailed example on running rqpd in r ... i do need it urgent plz ... many thanks in advance

The mata issues with qregpd ": 3499 mm_panels() not found" can be solved by installing moremata (just type: ssc install moremata). This will insure the ado finds the mata functions it needs.

You should try Unconditional Quantile Regression [ Firpo et al ( 2009 )] with fixed effects.

Fixed effects in unconditional quantile regression by N. T. Borgen.

https://med.mahidol.ac.th/ceb/sites/default/files/public/pdf/stata_journal/sj16-2.pdf ( page 403 )

Powell's 'genqreg' or 'qregpd' do not allow additive fixed effects and hence cannot be said to be a true 'fixed effects model' .