[FernandoRios]

Taruntej Singh : My rule of thumb is to use the method you understand. I do not know what qregpd does behind the blackbox, so I would not use it.
I understand, what xtqreg does, so i would be more comfortable taking about advantages and limitations of this estimator.

Hi Mia Liu
what you report is unexpected. I wonder if this has something to do with "singletons".
My own interpretation of what xtqreg does (which i called mmqreg and you can get it from SSC) actually drops the singletons.

Perhaps there is something else in your data that is causing this problem.
HTH
Fernando

[Joao Santos Silva]

Dear Mia Liu,

How are you defining the panel for xtqreg?

Best wishes,

Joao

[rabiya gill]

My humble question is that Fixed effect panel quantile regression can control the cross sectional dependence or not? If cross sectional dependence exists in panel data can we use XTQREG?



Regards,
Rabiya

[OWUSU ANSAH]

Hi, Joao do we have a command to plot the coefficients after running xtqreg?
Thankyou

[FernandoRios]

Hi Owusu,
There is no command to get plots of the coefficients, however, you could adapt the code I posted here, changing uqreg with xtqreg, and changing i=5(5)95 to i=0.05(.05).95
HTH

[OWUSU ANSAH]

Hi Owusu,
There is no command to get plots of the coefficients, however, you could adapt the code I posted here, changing uqreg with xtqreg, and changing i=5(5)95 to i=0.05(.05).95
HTH

Thankyou for this I appreciate it a lot, however I am finding it difficult to implement this with my variables. My model is below, can you please do me a favour and elaborate more on this code for me.
xtqreg gdppc poplog hc fdi lly lly_fdi i. yearavg5 , i(countryID) ls q(0.25)
eststo
xtqreg gdppc poplog hc fdi lly lly_fdi i. yearavg5 , i(countryID) ls q(0.5)
eststo
xtqreg gdppc poplog hc fdi lly lly_fdi i. yearavg5, i(countryID) ls q(0.75)
eststo
xtqreg gdppc poplog hc fdi lly lly_fdi i. yearavg5, i(countryID) ls q(0.95)
eststo

Thankyou.

[Mia Liu]

Thank you FernandoRios and Joao Santos Silva. I may have figured out why my number of observations is smaller when using xtqreg. When I include "nodrop" in the bootstrap option, then my number of observations is the same as using ols. Would you think that may be the reason?

Thank you,
Mia

[Joao Santos Silva]

Dear Mia Liu,

It looks as if you are not defining your sub-samples correctly. I suggest that you write a simplified version of your code that can replicate the problem, and post it and the results here for us to look at.

Best wishes,

Joao

[OWUSU ANSAH]

Hello I am using this code to plot the coefficient after quantile regression. Can someone help me how to plot multiple coefficients after running the regression.
this code plots only one coefficient at a time. Moreover how do I change the quantile to 0.25,0.5,0.75,0.95


Thankyou

code

tempfile holding

postfile handle quantile coefficient se using `holding'

forvalues i=1(1)9{

local ii=`i'/10

qregpd gdppc gdppc_lag fdi poplog hc afdi afdi_fdi,quantile(0.5) id(countrycode) fix(yearavg5) optimize(mcmc) noisy draws(1000) burn(100) arate(.5)

post handle (`ii') (_b[fdi]) (_se[fdi])

}

postclose handle

use `holding', clear

gen lb=coefficient-se*1.65

gen ub=coefficient+se*1.65

graph twoway (rarea lb ub quantile) || (line coefficient quantile)

[Yen Ng]

Dear Cobby Stoneson,

T=5 is too short for xtqreg. You can try it , but I would go for the "correlated fixed effects" (or Mundlak device) mentioned in the posts above.

Best wishes,

Joao

Hi Joao,
I have a panel of 230k obs in 9 years. Is it possible to use xtqreg? Thank you.

Best regards,
Yen

[Joao Santos Silva]

Dear Yen Ng,

I would say that T=9 is below what is desirable, put it is probably worth a try. You should, however, use the jackknife correction.

Best wishes,

Joao

[Yen Ng]

Dear Yen Ng,

I would say that T=9 is below what is desirable, put it is probably worth a try. You should, however, use the jackknife correction.

Best wishes,

Joao

Thanks Joao.

[Mirian Wawrzyniak]

Dear Joao Santos Silva

It is possible to use xtqreg when my dependent variable is counting and inflated to zero?
My database is an unbalanced panel with approximately 2800 firms and 20 years.

Best regards,
Mirian

[Joao Santos Silva]

Dear Mirian Wawrzyniak,

I am afraid it does not work for count data. For counts you can use qcount but that does not allow fixed effects.

Best wishes,

Joao

[Elif Cengen]

@Joao Santos Silva, many thanks for all of the knowledge above.

In one of the posts, you mentioned that xtqreg is not proper for short time periods. What is the plausible number of years to be used with the command? I have panel data capturing two-years , but my coeficients differ across different quantiles when I estimate my model with xtqreg . Is it still not proper to run the command with two-years of time time span? Best regards.