[Siddharth Kaushal]

Dear Joao Santos Silva

Thank you for your guidance. I have shifted towards using the "correlated random effects" approach. Apologies for the delay in responding.

Regards,
Siddharth

[Hachemi SOLTANI]

Thanks a lot, Mr. Joao Santos Silva

Can the xtqreg command be applied to a sample of 12 banks over 13 quarters (ie 156 observations in total)

[Joao Santos Silva]

Dear Hachemi SOLTANI

That is probably the bare minimum number of observations you need, so be careful when interpreting the results.

Best wishes,

Joao

[Atif Awad]

Dear Silva
greeting
Currently, I'm working on analyzing panel data using xtqreg. Although several published papers used MM-QR regression, there is no interpretation for Location and Scale parameters. Your help in interpreting these coefficients is highly appreciated.

xtqreg dlefp dleci dlre dly dlur dw , i( id )	quantile(.1(0.1)0.9) ls
MM-QR regression	results
Number of obs = 600
	Location parameters
(Std. err.	adjusted for 24 clusters in id)
Robust
dlefp Coefficient std. err. t	P>t [95% conf. interval]
dleci .6745461 .1660505 4.06	0.000 .3310444 1.018048
dlre .0093567 .0706979 0.13	0.896 -.136893 .1556064
dly .4348088 .0726831 5.98	0.000 .2844524 .5851652
dlur .2038351 .1794185 1.14	0.268 -.1673204 .5749906
dw -.1690325 .0435712 -3.88	0.001 -.2591664 -.0788986
_cons -.0155811 .0069295 -2.25	0.034 -.0299159 -.0012464
Scale parameters
(Std. err.	adjusted for 24 clusters in id)
Robust
Coefficient std. err. t	P>t [95% conf. interval]
dleci .0909063 .1219005 0.75	0.463 -.1612641 .3430766
dlre .0444691 .0422671 1.05	0.304 -.0429671 .1319053
dly .0307853 .0704269 0.44	0.666 -.1149038 .1764745
dlur .126084 .0791408 1.59	0.125 -.0376312 .2897991
dw -.0189435 .03067 -0.62	0.543 -.0823891 .0445022
_cons .0389274 .0033464 11.63	0.000 .0320049 .0458499
.1 Quantile regression
Coefficient Std. err. z	P>z [95% conf. interval]
dleci .5416918 .5242467 1.03	0.301 -.4858128 1.569196
dlre -.0556323 .1710175 -0.33	0.745 -.3908204 .2795558
dly .3898178 .1303774 2.99	0.003 .1342828 .6453527
dlur .0195706 .221195 0.09	0.929 -.4139636 .4531049
dw -.1413477 .1236448 -1.14	0.253 -.383687 .1009916
.2 Quantile regression
Coefficient Std. err. z	P>z [95% conf. interval]
dleci .585722 .411174 1.42	0.154 -.2201642 1.391608
dlre -.0340938 .1341357 -0.25	0.799 -.2969949 .2288072
dly .4047286 .1022569 3.96	0.000 .2043087 .6051484
dlur .080639 .1734987 0.46	0.642 -.2594123 .4206903
dw -.1505229 .0969757 -1.55	0.121 -.3405918 .039546
.3 Quantile regression
Coefficient Std. err. z	P>z [95% conf. interval]
dleci .6169928 .3504955 1.76	0.078 -.0699658 1.303951
dlre -.0187969 .1143441 -0.16	0.869 -.2429072 .2053134
dly .4153184 .087167 4.76	0.000 .2444742 .5861626
dlur .1240106 .1479111 0.84	0.402 -.1658899 .4139111
dw -.1570393 .0826642 -1.90	0.057 -.3190581 .0049796
.4 Quantile regression
Coefficient Std. err. z	P>z [95% conf. interval]
dleci .6424482 .3208702 2.00	0.045 .0135541 1.271342
dlre -.0063448 .104686 -0.06	0.952 -.2115255 .198836
dly .4239389 .0798014 5.31	0.000 .267531 .5803467
dlur .1593164 .1354559 1.18	0.240 -.1061723 .4248052
dw -.1623438 .0756762 -2.15	0.032 -.3106664 -.0140211
.5 Quantile regression
Coefficient Std. err. z	P>z [95% conf. interval]
dleci .6701909 .315267 2.13	0.034 .052279 1.288103
dlre .0072263 .1028611 0.07	0.944 -.1943778 .2088303
dly .4333339 .0784089 5.53	0.000 .2796552 .5870126
dlur .1977945 .133115 1.49	0.137 -.0631061 .4586952
dw -.1681249 .0743543 -2.26	0.024 -.3138566 -.0223932
.6 Quantile regression
Coefficient Std. err. z	P>z [95% conf. interval]
dleci .6965017 .3368394 2.07	0.039 .0363086 1.356695
dlre .0200969 .1098937 0.18	0.855 -.1952909 .2354846
dly .442244 .0837723 5.28	0.000 .2780534 .6064347
dlur .2342868 .1421818 1.65	0.099 -.0443844 .512958
dw -.1736077 .0794428 -2.19	0.029 -.3293127 -.0179027
.7 Quantile regression
Coefficient Std. err. z	P>z [95% conf. interval]
dleci .7268043 .388228 1.87	0.061 -.0341087 1.487717
dlre .0349201 .1266516 0.28	0.783 -.2133125 .2831528
dly .452506 .0965504 4.69	0.000 .2632707 .6417412
dlur .2763155 .1638204 1.69	0.092 -.0447667 .5973976
dw -.1799223 .0915637 -1.96	0.049 -.3593838 -.0004608
.8 Quantile regression
Coefficient Std. err. z	P>z [95% conf. interval]
dleci .7593515 .4646091 1.63	0.102 -.1512656 1.669969
dlre .0508415 .151567 0.34	0.737 -.2462244 .3479073
dly .4635281 .1155455 4.01	0.000 .2370631 .6899931
dlur .3214574 .196038 1.64	0.101 -.0627699 .7056848
dw -.1867046 .1095785 -1.70	0.088 -.4014746 .0280653
.9 Quantile regression
Coefficient Std. err. z	P>z [95% conf. interval]
dleci .8197206 .6368689 1.29	0.198 -.4285195 2.067961
dlre .0803725 .2077761 0.39	0.699 -.3268612 .4876063
dly .4839721 .1583916 3.06	0.002 .1735303 .7944138
dlur .4051874 .2688455 1.51	0.132 -.1217401 .9321149
dw -.1992846 .1502044 -1.33	0.185 -.4936798 .0951105
.

[Joao Santos Silva]

Dear Atif Awad,

The location parameters are the ones you would obtain with xtreg, so they are just the parameters of the conditional mean. The scale parameters show how the variables affect the conditional dispersion of the dependent variable; if these parameters were all zero (except for the constant) all the quantiles would have the same slope. The original paper has examples.

Best wishes,

Joao

[valeria santa maria]

Dear Joao Santos Silva

I had a question about the XTQREG command. I am currently doing a quantile regression; however, I want to see if the model has heteroscedasticity, autocorrelation, and multicollinearity problems. I wanted to know which would be the command that should be used since the ones found in Stata are for linear regressions but not for quantiles.

I thank you in advance for your reply.

[valeria santa maria]

Dear Joao Santos Silva

I had a question about the XTQREG command. I am currently doing a quantile regression; however, I want to see if the model has heteroscedasticity, autocorrelation, and multicollinearity problems. I wanted to know which would be the command that should be used since the ones found in Stata are for linear regressions but not for quantiles.

I thank you in advance for your reply.

[Joao Santos Silva]

Dear valeria santa maria,

You almost certainly have heteroskedasticity and serial correlation, but is you compute the standard errors by clustered bootstrap as in this example, you do not have to worry about that. Multicollinearity you will have for sure, but that is always the case, so it is not different in this context. Finally, I do not understand what you say about the commands because xtqreg is for quantile regression.

Best wishes,

Joao

[valeria santa maria]

Thank you Joao. I would like to know the use of this command and if I need to change the number according to my regression? or is it standard?

[Joao Santos Silva]

Dear valeria santa maria,

The bit about c>=10 is specific to that example and should be dropped. Apart from that, you should just change the name of the variables and increase the number of replicas.

Best wishes,

Joao

[Panagiotis Palaios]

Dear Justin Matz,

As noted by Fernando, both mmqreg and xtqreg implement essentially the same estimator, the one proposed in

Machado, J.A.F. and Santos Silva, J.M.C. (2019), Quantiles via Moments, Journal of Econometrics, 213(1), pp. 145–173.

There are minor differences between the two commands (e.g., xtqreg always includes one set of fixed effects, while by default mmqreg does not include fixed effects) but the results should be the same when you use the same specification in both commands. For details on the commands and the method, please check the help files and the paper mentioned above.

Best wishes,

Joao

Dear Joao Santos Silva , FernandoRios

As always your comments are very helpful.
I am trying to apply the denopt(denmethod bwidth) option in mmqreg command, but I get an "invalid" message:

The command is the following:

mmqreg GDPpc_g temp temp_2, q(10 20 30 40 50 60 70 80 90) robust denopt(res bo)

and the message I get is:
invalid 'res'

or, If i try another specification:

mmqreg GDPpc_g temp temp_2, q(10 20 30 40 50 60 70 80 90) robust denopt(fitted bofinger)

the message I get is:
invalid 'fitted'

Could you please help me to clarify the syntax of that command, namely denopt?

Thank you in advance,
Panagiotis

[FernandoRios]

two questions
1. do you have the latest mmqreg iteration?
2. what Stata version you are using?
I just run similar commands on my end, and i get no error, so perhaps is something related to your setup

[valeria santa maria]

Dear valeria santa maria,

The bit about c>=10 is specific to that example and should be dropped. Apart from that, you should just change the name of the variables and increase the number of replicas.

Best wishes,

Joao

Dear Joao Santos Silva

I would like to know if there is another command to solve the heteroscedasticity problem with a quantile regression because I am trying the JK correction and do not work for me.

I thank you in advance for your reply.

[Joao Santos Silva]

Dear valeria santa maria,

There is some confusion here because there is no heteroskedacity problem with quantile regression. Also, depending on your problem, the JK correction may not be needed. Anyway, I may be able to provide more help if you explain why the correction is not working.

Best wishes,

Joao

[valeria santa maria]

Dear Joao Santos Silva

Currently I am investigating factor that influence the ecological footprint in Latin America. The model we are using for this paper is the quantile regression model. We are using the XTQREG command.
My question is if I use that command, the results that come out are valid or I have to correct the model?

I thank you in advance for your reply.

Valeria.