how to tabulate results from quantile regression

[Ilias Geo]

Dear all,

I was wondering how someone can tabulate results from a quantile regression. Below is a sample of what I am doing. Variables Rown, Dlab1 Dlab2 Deduc1 are dummy vars, nwealth, emo inher are in thousands and size/ra0300 are integers. I would like to make a quantile regression table showing coefficients, std.errors and significance levels (1%, 5%, 10%) for quantiles .1 .25 .5 .75 and 0.9

I used eststo: mi estimate:.. but it returns no results when tried to display them.

Any help would be much appreciated.
Ilias


. mi estimate: sqreg nwealth emo inher size ra0300 Rown Dlab1 Dlab2 Deduc1, quantile(.1 .25 .5 .75 .9) reps(500)
xxxxxxxxx
Multiple-imputation estimates Imputations = 5
Simultaneous quantile regression Number of obs = 204
Average RVI = 0.1528
Largest FMI = 0.2762
Complete DF = 195
DF adjustment: Small sample DF: min = 43.96
avg = 136.69
max = 192.98

------------------------------------------------------------------------------
nwealth | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
q10 |
emo | 6.514076 1.452832 4.48 0.000 3.64837 9.379782
inher | .7629998 .2970967 2.57 0.011 .1770264 1.348973
size | 21523.16 11910.8 1.81 0.073 -2027.945 45074.27
ra0300 | 3080.98 1513.365 2.04 0.044 85.20319 6076.757
Rown | 176529.4 34403.11 5.13 0.000 108630.2 244428.6
Dlab1 | 79280.07 77806.56 1.02 0.310 -74194.63 232754.8
Dlab2 | 7229.486 27602.12 0.26 0.794 -48054.87 62513.84
Deduc1 | 62947.28 53146.36 1.18 0.238 -41892.48 167787
_cons | -387149.1 137062.8 -2.82 0.005 -658165.9 -116132.3
-------------+----------------------------------------------------------------

[Milla Nyyssola]

Hi Ilias,

I'm happy I found your post. Sqreg seems to be a handy tool in inserting all five deciles in one table (unfortunately it does not seem to allow the cluster() option?).
I used an outreg2 command on a sqreg similar to yours below on my data and I was able to create a table with coefficients, std.errors and significance levels (1%, 5%, 10%) for quantiles .1 .25 .5 .75 and 0.9 in Excel. If you need other file types, just type the needed extension.

sqreg lnhhinc97N_pae_inf treat inacshare hhsize dep femhead if nvals==1 & samp98==1, quantile(.1 .25 .5 .75 .9) reps(500)
outreg2 using hmhm.xls, bdec(5) replace

The table looks like this:

	(1)	(2)	(3)	(4)	(5)
VARIABLES	q10	q25	q50	q75	q90
treat	-0.39501***	-0.07104***	-0.03525***	-0.01699	-0.03815**
	(0.047)	(0.016)	(0.010)	(0.013)	(0.018)
inacshare	-4.11437***	-1.50727***	-0.62698***	-0.30552	0.30679**
	(0.382)	(0.320)	(0.154)	(0.193)	(0.142)
hhsize	-0.05245***	-0.06806***	-0.03993***	-0.03050***	-0.02940***
	(0.015)	(0.005)	(0.003)	(0.004)	(0.003)
dep	-0.68831***	-0.83820***	-1.05511***	-1.09529***	-1.01679***
	(0.165)	(0.043)	(0.041)	(0.039)	(0.054)
femhead	-0.11261	0.01970	0.21009***	0.18106***	0.17015***
	(0.072)	(0.031)	(0.020)	(0.023)	(0.026)
Constant	5.07105***	5.72221***	6.06719***	6.49233***	6.92404***
	(0.091)	(0.026)	(0.020)	(0.017)	(0.026)
Observations	20,047	20,047	20,047	20,047	20,047
Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1

Best,
Milla

[Ilias Geo]

Hi Milla,

Thanks for your reply and your suggestion.

I had to modify my mi estimate command to this: mi estimate, dots post: sqreg nwealth emo inher size ra0300 Rown Dlab1 Dlab2 Deduc1, quantile(.1 .25 .5 .75 .9) reps(500) in order for the outreg2 to work. Before the modification I was getting the following message:

matrix e(b) not found; run/post a regression, or specify varlist for non-regression outputs
r(111);

As you said, the cluster option is not allowed but you can overcome the problem by defining your own program.


Best,
Ilias