[Raymond Zhang]

So does it mean --mmqreg-- can only get heteroscedasticity robust standard error,while --xtqreg-- get cluster and robust error.because xtreg,fe vce(robust) get cluster and robust error,but --reghdfe,a(idcode) vce(robust)--only get heteroscedasticity robust standard error.Is it right?

Best
Raymond

[Raymond Zhang]

I find that the two command you show in #15 get the same standard error.

  xtqreg ln_w age c.age#c.age ttl_exp c.ttl_exp#c.ttl_exp tenure c.tenure#c.tenure not_smsa south, i(idcode) ls  
xtreg ln_w age c.age#c.age ttl_exp c.ttl_exp#c.ttl_exp tenure c.tenure#c.tenure not_smsa south, fe vce(robust)

Because these two command get the absolutely the same standard error, So can i conclude that the standard error in --xtqreg-- is cluster and robust error?

[FernandoRios]

No, this just means that -xtqreg- reports the standard errors from -xtreg, robust- when you request seeing the location - scale regressions.
My command -mmqreg- obtains the standard errors following their paper. So it is constructed under different assumptions than xtreg. Specifically, when robust is not added, standard errors are constructed assuming a particular type of heteroskedasticity. (the one modelled in the scale regression)
HTH
Fernando

[Zeeshan Fareed]

How to plot quantile coefficients obtained from mmqreg? (plotting coefficients against quantiles after mmqreg for each variable)

[FernandoRios]

You will need to things.
First, change mmqreg with the file im attaching (its an update that will soon send to SSC), and then download from ssc "qregplot"
Let me know if you have problems.

[Zeeshan Fareed]

@FernandoRios Much appreciated. Thank you so much. It works.

[Hamid muili]

Zeeshan Fareed can you pls share the code you used with the qregplot tto plot your mmqreg?

[Daniel Lasso]

dear FernandoRios

Thank you in advance. As you told me, I used mmqreg, it worked and has the same coefficients that the xtqreg model. Actually, I am just an undergraduate student trying to understand the methodology, therefore, i would really apreciate if you can explain to me what is the differene between location, scale and qtile results on the output of the mmqreg command. Also, may i ask, why does it have 3 constant?

Thank you,
D.

[FernandoRios]

Hi Daniel,
The method of moments proposed by Machado and Santos-Silva (2019) basically suggest that quantile regressions models can be estimated using 3 equations.
A location model (which models the conditional mean)
a Scale model (which models heteroskedasticity)
and a quantile (which models the position within a distribution)

This is done in a 4 step proccess
1. Estimate the location model using OLS.
y = b_0 + b_1 *x + e
2. Get the residuals and estimate the scale model (model of variance)
ABS(e) = g_0 + g_1*x + v
3. Get Standardized errors and get a particular quantile Q
ST_err = e / (g_0 + g_1*x)
and get q_t

The QTILE equation, basically displays the quantile regression coefficients that combine all of this:
QTILE = b_0 + g_0 * q_t

Let me know if this helps.

[Daniel Lasso]

FernandoRios

I understand now, thank you so much!

[Anuradha Saikia]

Hello everyone. I am getting this error while executing the following

Also I am getting this error.
code:
mmqreg ROA c.DTTA#i.InsDistancedummy LEV HHI GDPpercapita Size Age, abs(Companynum)

_i___0b__InsDistancedummy_X_c__DT invalid name
st_addvar(): 3300 argument out of range
map_solve(): - function returned error
<istmt>: - function returned error

[Anuradha Saikia]

Please help with the above.

Regards
Anuradha

[FernandoRios]

The name of the variable is too long
change them to something smaller and try again

[Anuradha Saikia]

okay FernandoRios . here DTTA is continuous and Instdistancedummy is categorical. is there an error with the specification as
c.DTTA and i.Instdistancedummy

[FernandoRios]

It’s the variable name
it’s too long for how mmqreg treats interactions
change the name to something shorter and try again