[Joao Santos Silva]

Dear Elif Cengen,

Two periods is certainly not enough and I am surprised that the estimates vary across quantiles. Can you please contact me by email so that we can explore what is going on? I can then post the finding here.

Best wishes,

Joao

[Michael Haylock]

Hi all,
According to the help file, one can access the coefficient vectors and covariance matrices from running an estimation with multiple quantiles simulatanously, i.e. with e(b_90) for the 9th decile point estimate. This does not seem to be the case (see results below). Or what am I doing wrong? The stored vector is empty (showed below). If this could be fixed, numerous problems could be solved that I am running into (elucidated here): This seems to be a general problem when estimating anything with more than one quantile, although in the help file, it is stated that one can do this.

" If multiple quantiles are estimated, xtqreg also saves b, V, and q for each quantile with names of the form (for the first quartile) e(b_25), e(V_25), and e(q_25). In this case, the matrices e(b), e(V), and e(q) contain the results for the last estimated quantile."

For example:

When estimating multiple quantiles, I cannot access the coefficient vector of anything apart from the last estimated quantile. Why is this the case? This is making my life a bit hard trying to make tables and do bootstrapping while estimaing quantiles simultaneously. Is there a way around this?
Below is code and results, not bootstrapping for time's sake (commented out), but to show the problem I am having. Otherwise I have to estimate each quantile separately with a bootstrap that takes a long time for me.

Also, it would be fantastic if when i store results using esttab or eststo: xtqreg .... with the "ls" option, that the location and scale functions can be automatically tabulated as well. At the moment I can only get the Quantile point estimate in a table automatically (esttab after storing results only gives the point estimate and not location and scale functions) and have to make workarounds that make it somewhat tedious to get a publishable table in due time. If anyone has a solution here I would be very thankful.

Ideally, after estimating multiple quantiles simultaneously, e.g.

#
bs, cluster(id) idcluster(newid) group(id) rep(200): xtqreg logTotalWealth MV_shock_sc MV_trend_sc IND_MV_shock_sc IND_MV_trend_sc GDP_growth imf24 imf33 i.Year Age Age_sq, id(id) ls q(0.1(0.2)0.9)

and then running esttab after saving, I should get location, scale, and the point estimates all in one table, and these should be saved.
But if I do this, it would only give the last estimated quantile, and no other results. This seems to be related to the above issue of only accessing the last estimated quantile point estimate (here 0.9) when estimating mulitple quantiles. Also, then I could do a JK Bias correction after saving the different b_quantile vectors, covariance matrices, and quantile errors from running an estimation with multiple quantiles.
The main point being, I then only need to run one bootstrap loop which already takes a long time, and not one for each quantile, and then no longer copy and paste my results...

Thanks in advance,
Michael

[Michael Haylock]

Post deleted

[Michael Haylock]

. /*bs, cluster(id) idcluster(newid) group(id) rep(200):*/ xtqreg logTotalCompensation MV_shock_sc MV_trend_sc IND_MV_shock_sc IND_MV_trend_sc GDP_growth imf24 imf33 i.Year Age Age_sq
> , id(id) ls quantile(0.1(0.2)0.9)



MM-QR regression results
Number of obs = 25706

Location parameters
(Std. Err. adjusted for 6,946 clusters in id)
---------------------------------------------------------------------------------
| Robust
logTotalCompe~n | Coef. Std. Err. t P>|t| [95% Conf. Interval]
----------------+----------------------------------------------------------------
MV_shock_sc | .0046484 .0012089 3.85 0.000 .0022785 .0070184
MV_trend_sc | .0050125 .0021142 2.37 0.018 .0008681 .009157
IND_MV_shock_sc | .0007126 .0021668 0.33 0.742 -.0035349 .0049602
IND_MV_trend_sc | .0014187 .0020131 0.70 0.481 -.0025277 .005365
GDP_growth | .010524 .0071083 1.48 0.139 -.0034104 .0244584
imf24 | .0000215 3.04e-06 7.08 0.000 .0000156 .0000275
imf33 | .0246066 .0105961 2.32 0.020 .0038351 .0453782
|
Year |
2003 | .1759501 .1707837 1.03 0.303 -.1588381 .5107382
2004 | .0283583 .1895973 0.15 0.881 -.3433103 .4000269
2005 | .2930514 .2371676 1.24 0.217 -.1718695 .7579723
2006 | .5028077 .2801 1.80 0.073 -.0462739 1.051889
2007 | .6271276 .3280158 1.91 0.056 -.0158837 1.270139
2008 | .4645995 .3781433 1.23 0.219 -.2766771 1.205876
2009 | .8845965 .4259698 2.08 0.038 .0495654 1.719628
2010 | .9456714 .4742507 1.99 0.046 .0159951 1.875348
2011 | .9454018 .5287649 1.79 0.074 -.091139 1.981943
2012 | 1.115021 .5797992 1.92 0.055 -.0215626 2.251605
2013 | 1.282268 .6343921 2.02 0.043 .0386653 2.52587
|
Age | .1296837 .059011 2.20 0.028 .0140041 .2453634
Age_sq | -.0018104 .0002101 -8.62 0.000 -.0022223 -.0013985
_cons | 3.142591 2.537942 1.24 0.216 -1.832551 8.117732
---------------------------------------------------------------------------------
Scale parameters
(Std. Err. adjusted for 6,946 clusters in id)
---------------------------------------------------------------------------------
| Robust
| Coef. Std. Err. t P>|t| [95% Conf. Interval]
----------------+----------------------------------------------------------------
MV_shock_sc | -.0014445 .0005711 -2.53 0.011 -.0025641 -.0003249
MV_trend_sc | .0024466 .0009685 2.53 0.012 .000548 .0043452
IND_MV_shock_sc | -.0008285 .0010047 -0.82 0.410 -.0027979 .001141
IND_MV_trend_sc | -.0010953 .0009218 -1.19 0.235 -.0029023 .0007117
GDP_growth | -.0061733 .0032435 -1.90 0.057 -.0125316 .0001849
imf24 | 2.87e-06 1.45e-06 1.98 0.048 2.49e-08 5.71e-06
imf33 | .0069888 .0048692 1.44 0.151 -.0025563 .016534
|
Year |
2003 | .1694704 .0988755 1.71 0.087 -.0243558 .3632967
2004 | .0225159 .0746856 0.30 0.763 -.1238908 .1689225
2005 | .1613661 .122404 1.32 0.187 -.0785832 .4013153
2006 | .1399702 .1411914 0.99 0.322 -.1368081 .4167485
2007 | .0983096 .1628371 0.60 0.546 -.2209008 .41752
2008 | .0389999 .1855849 0.21 0.834 -.3248033 .402803
2009 | .0453703 .2072373 0.22 0.827 -.3608781 .4516187
2010 | .0520061 .2288973 0.23 0.820 -.3967026 .5007149
2011 | .0061091 .2540092 0.02 0.981 -.4918267 .5040448
2012 | .0059601 .278116 0.02 0.983 -.5392322 .5511524
2013 | .0281614 .3025727 0.09 0.926 -.5649737 .6212964
|
Age | -.0444984 .0275973 -1.61 0.107 -.0985976 .0096008
Age_sq | .0004467 .0000967 4.62 0.000 .0002571 .0006363
_cons | 1.175088 1.192076 0.99 0.324 -1.161745 3.511921
---------------------------------------------------------------------------------

WARNING: some fitted values of the scale function are negative
.1 Quantile regression
---------------------------------------------------------------------------------
| Coef. Std. Err. z P>|z| [95% Conf. Interval]
----------------+----------------------------------------------------------------
MV_shock_sc | .006878 .0145377 0.47 0.636 -.0216155 .0353714
MV_trend_sc | .0012363 .0184588 0.07 0.947 -.0349423 .0374148
IND_MV_shock_sc | .0019914 .0236675 0.08 0.933 -.0443962 .0483789
IND_MV_trend_sc | .0031092 .0203877 0.15 0.879 -.0368499 .0430683
GDP_growth | .0200524 .0794101 0.25 0.801 -.1355886 .1756933
imf24 | .0000171 .0000292 0.59 0.558 -.0000401 .0000743
imf33 | .0138196 .1088064 0.13 0.899 -.1994371 .2270762
|
Year |
2003 | -.0856229 2.414779 -0.04 0.972 -4.818504 4.647258
2004 | -.0063943 2.781438 -0.00 0.998 -5.457912 5.445124
2005 | .0439873 3.247157 0.01 0.989 -6.320324 6.408298
2006 | .2867675 3.608065 0.08 0.937 -6.784909 7.358444
2007 | .4753894 4.033332 0.12 0.906 -7.429797 8.380575
2008 | .4044042 4.486745 0.09 0.928 -8.389454 9.198262
2009 | .8145687 4.937671 0.16 0.869 -8.863088 10.49223
2010 | .8654014 5.41444 0.16 0.873 -9.746707 11.47751
2011 | .9359726 5.945422 0.16 0.875 -10.71684 12.58879
2012 | 1.105822 6.454042 0.17 0.864 -11.54387 13.75551
2013 | 1.238801 6.9955 0.18 0.859 -12.47213 14.94973
|
Age | .1983658 .609973 0.33 0.745 -.9971592 1.393891
Age_sq | -.0024998 .0018698 -1.34 0.181 -.0061646 .0011649
---------------------------------------------------------------------------------

.3 Quantile regression
---------------------------------------------------------------------------------
| Coef. Std. Err. z P>|z| [95% Conf. Interval]
----------------+----------------------------------------------------------------
MV_shock_sc | .0058171 .0101092 0.58 0.565 -.0139966 .0256308
MV_trend_sc | .0030331 .0128358 0.24 0.813 -.0221247 .0281908
IND_MV_shock_sc | .0013829 .0164578 0.08 0.933 -.0308738 .0336397
IND_MV_trend_sc | .0023048 .0141771 0.16 0.871 -.0254818 .0300915
GDP_growth | .0155187 .0552199 0.28 0.779 -.0927104 .1237477
imf24 | .0000192 .0000203 0.95 0.344 -.0000206 .000059
imf33 | .0189522 .0756614 0.25 0.802 -.1293415 .1672458
|
Year |
2003 | .0388366 1.67918 0.02 0.982 -3.252296 3.329969
2004 | .0101414 1.934145 0.01 0.996 -3.780713 3.800995
2005 | .1624949 2.257995 0.07 0.943 -4.263095 4.588085
2006 | .3895619 2.508962 0.16 0.877 -4.527912 5.307036
2007 | .5475882 2.804682 0.20 0.845 -4.949488 6.044664
2008 | .4330458 3.119974 0.14 0.890 -5.681991 6.548083
2009 | .8478888 3.433537 0.25 0.805 -5.88172 7.577498
2010 | .9035948 3.765071 0.24 0.810 -6.475809 8.282999
2011 | .9404591 4.134303 0.23 0.820 -7.162626 9.043545
2012 | 1.110199 4.487985 0.25 0.805 -7.68609 9.906488
2013 | 1.259483 4.864502 0.26 0.796 -8.274766 10.79373
|
Age | .1656861 .4241608 0.39 0.696 -.6656539 .997026
Age_sq | -.0021718 .0013002 -1.67 0.095 -.0047202 .0003766
---------------------------------------------------------------------------------

.5 Quantile regression
---------------------------------------------------------------------------------
| Coef. Std. Err. z P>|z| [95% Conf. Interval]
----------------+----------------------------------------------------------------
MV_shock_sc | .0045895 .0104258 0.44 0.660 -.0158447 .0250236
MV_trend_sc | .0051125 .0132378 0.39 0.699 -.0208332 .0310581
IND_MV_shock_sc | .0006788 .0169732 0.04 0.968 -.032588 .0339456
IND_MV_trend_sc | .0013739 .014621 0.09 0.925 -.0272828 .0300307
GDP_growth | .0102719 .0569491 0.18 0.857 -.1013462 .12189
imf24 | .0000217 .0000209 1.03 0.301 -.0000194 .0000627
imf33 | .0248921 .0780306 0.32 0.750 -.1280451 .1778293
|
Year |
2003 | .1828716 1.731761 0.11 0.916 -3.211318 3.577061
2004 | .0292779 1.994708 0.01 0.988 -3.880278 3.938834
2005 | .299642 2.3287 0.13 0.898 -4.264527 4.863811
2006 | .5085244 2.587525 0.20 0.844 -4.562931 5.579979
2007 | .6311428 2.892505 0.22 0.827 -5.038062 6.300348
2008 | .4661923 3.217669 0.14 0.885 -5.840323 6.772707
2009 | .8864495 3.54105 0.25 0.802 -6.053882 7.826781
2010 | .9477955 3.882966 0.24 0.807 -6.662678 8.558269
2011 | .9456513 4.263759 0.22 0.824 -7.411163 9.302466
2012 | 1.115264 4.628516 0.24 0.810 -7.95646 10.18699
2013 | 1.283418 5.016823 0.26 0.798 -8.549374 11.11621
|
Age | .1278663 .4374429 0.29 0.770 -.729506 .9852386
Age_sq | -.0017921 .001341 -1.34 0.181 -.0044204 .0008361
---------------------------------------------------------------------------------

.7 Quantile regression
---------------------------------------------------------------------------------
| Coef. Std. Err. z P>|z| [95% Conf. Interval]
----------------+----------------------------------------------------------------
MV_shock_sc | .0035007 .0152866 0.23 0.819 -.0264604 .0334619
MV_trend_sc | .0069565 .0194096 0.36 0.720 -.0310856 .0449985
IND_MV_shock_sc | .0000544 .0248866 0.00 0.998 -.0487226 .0488313
IND_MV_trend_sc | .0005484 .0214379 0.03 0.980 -.041469 .0425658
GDP_growth | .005619 .0835005 0.07 0.946 -.1580389 .1692769
imf24 | .0000238 .0000307 0.78 0.438 -.0000364 .000084
imf33 | .0301596 .114411 0.26 0.792 -.1940818 .2544011
|
Year |
2003 | .3106028 2.539165 0.12 0.903 -4.666069 5.287274
2004 | .0462483 2.92471 0.02 0.987 -5.686077 5.778574
2005 | .4212648 3.414418 0.12 0.902 -6.270871 7.113401
2006 | .614021 3.793916 0.16 0.871 -6.821917 8.049959
2007 | .7052395 4.241089 0.17 0.868 -7.607142 9.017621
2008 | .4955868 4.717857 0.11 0.916 -8.751243 9.742416
2009 | .9206454 5.19201 0.18 0.859 -9.255507 11.0968
2010 | .9869929 5.693338 0.17 0.862 -10.17174 12.14573
2011 | .9502558 6.251671 0.15 0.879 -11.30279 13.20331
2012 | 1.119757 6.786489 0.16 0.869 -12.18152 14.42103
2013 | 1.304643 7.355838 0.18 0.859 -13.11254 15.72182
|
Age | .0943275 .6413924 0.15 0.883 -1.162779 1.351434
Age_sq | -.0014554 .0019661 -0.74 0.459 -.005309 .0023981
---------------------------------------------------------------------------------

.9 Quantile regression
---------------------------------------------------------------------------------
| Coef. Std. Err. z P>|z| [95% Conf. Interval]
----------------+----------------------------------------------------------------
MV_shock_sc | .0025071 .0211082 0.12 0.905 -.0388642 .0438784
MV_trend_sc | .0086394 .0268014 0.32 0.747 -.0438903 .0611691
IND_MV_shock_sc | -.0005155 .0343643 -0.02 0.988 -.0678683 .0668373
IND_MV_trend_sc | -.000205 .0296021 -0.01 0.994 -.058224 .057814
GDP_growth | .0013726 .1153002 0.01 0.991 -.2246116 .2273568
imf24 | .0000258 .0000424 0.61 0.543 -.0000573 .0001089
imf33 | .0349671 .1579825 0.22 0.825 -.2746729 .344607
|
Year |
2003 | .4271768 3.506163 0.12 0.903 -6.444776 7.299129
2004 | .0617363 4.038536 0.02 0.988 -7.853649 7.977122
2005 | .532264 4.714741 0.11 0.910 -8.708458 9.772986
2006 | .7103026 5.238764 0.14 0.892 -9.557487 10.97809
2007 | .7728639 5.856236 0.13 0.895 -10.70515 12.25088
2008 | .5224137 6.514573 0.08 0.936 -12.24591 13.29074
2009 | .9518544 7.1693 0.13 0.894 -13.09971 15.00342
2010 | 1.022766 7.86155 0.13 0.896 -14.38559 16.43112
2011 | .954458 8.632514 0.11 0.912 -15.96496 17.87388
2012 | 1.123856 9.371009 0.12 0.905 -17.24298 19.4907
2013 | 1.324015 10.15719 0.13 0.896 -18.5837 21.23173
|
Age | .0637183 .8856558 0.07 0.943 -1.672135 1.799572
Age_sq | -.0011482 .0027149 -0.42 0.672 -.0064692 .0041729
---------------------------------------------------------------------------------


. mat b=e(b_location)

. mat g=e(b_scale)

. mat c=e(b_50)

. mat d=b,g,c

. eststo: ereturn post d
estimates post: matrix has missing values
r(504);

end of do-file

r(504);

. ssc install xtqreg, replace
checking xtqreg consistency and verifying not already installed...
all files already exist and are up to date.

. mat list c

symmetric c[1,1]
c1
r1 .


. mat z=e(b)

. mat list z

z[1,21]
2002b. 2003. 2004. 2005. 2006.
MV_shock_sc MV_trend_sc IND_MV_sho~c IND_MV_tre~c GDP_growth imf24 imf33 Year Year Year Year Year
y1 .00250712 .00863943 -.0005155 -.000205 .00137258 .00002579 .03496706 0 .42717677 .06173635 .53226403 .7103026

2007. 2008. 2009. 2010. 2011. 2012. 2013.
Year Year Year Year Year Year Year Age Age_sq
y1 .77286391 .52241372 .95185438 1.0227665 .95445801 1.1238564 1.3240147 .06371834 -.00114818

.

[Joao Santos Silva]

Dear Michael Haylock,

Please use ereturn post to see the saved results. I believe the estimates for the median are saved in e(b_5), not e(b_50).

Best wishes,

Joao

[Mohamed Mahjoub Elheddad]

Joao Santos Silva I would like to ask whether the Machado and Silva Quantile Regression mitigates the issue of endogeneity.
Thanks.

Mohamed

[Joao Santos Silva]

Dear Mohamed Mahjoub Elheddad,

There are 2 commands related to that paper: xtqreg allows for fixed effects and therefore deals with endogeneity only to the extent that the inclusion of the fixed effects can solve the endogeneity; ivqreg2 allows the estimation using instruments, but does not allow for fixed effects.

Best wishes,

Joao

[Mohamed Mahjoub Elheddad]

Dear Mohamed Mahjoub Elheddad,

There are 2 commands related to that paper: xtqreg allows for fixed effects and therefore deals with endogeneity only to the extent that the inclusion of the fixed effects can solve the endogeneity; ivqreg2 allows the estimation using instruments, but does not allow for fixed effects.

Best wishes,

Joao

Thanks very much for this elaboration. Very appreciated.

[Keith Dilts]

I am trying to test the equality of coefficients across quantiles--I know this can be done with bootstrapping, but I am unsure how to implement it. Would it be possible to request a very simple example?

[Joao Santos Silva]

Dear Keith Dilts

Indeed you can use bootstrap to do it, but a simpler approach is to check whether the variable of interest has a significant coefficient in the scale function; that is a test for the null hypothesis that the coefficient on that variables is the same across all quantiles.

Best wishes,

Joao

[OWUSU ANSAH]

I need help on xtqreg please. I keep getting different results results with the same command.

[OWUSU ANSAH]

I need help on xtqreg please. I keep getting different results results with the same command.

[FernandoRios]

THere is nothing wrong with your results.
You are requesting Bootstrap standard errors which depend on random draws. So every time you run it (unless you use "seeds") you will get different standard errors.

[OWUSU ANSAH]

THere is nothing wrong with your results.
You are requesting Bootstrap standard errors which depend on random draws. So every time you run it (unless you use "seeds") you will get different standard errors.

So i need to write 'seed'before my code? I mean this
seed
bootstrap, cluster(countrycode) idcluster(countryID) reps(100) : xtqreg gdppc lgdppc pop trade rir inf fdi dcpf i.yearavg5,i(countryID) ls q(0.9)

[Kyriaki Tselika]

Hello,

I am new to Stata so I am not familiar with many commands. I am using the xtqreg in my research and I was wondering if and how I could print the residuals.

Thank you in advance.

Kyriaki