Re: st: speeding up qreg

[Stas Kolenikov <skolenik@gmail.com>]

=== separate_qreg.do ===
args q
postfile topost double q coef using qreg_`q'.dta, replace
qreg cost group_a `list of fixed-effects' , quantile(`q')
post topost (`q') (_b[group_a])
=== EOF ===

clear
forval x=.1(.1).9 {
   ! stata -b do separate_qreg `q'
}
* wait one day
forval x=.1(.1).9 {
   append using qreg_`q'.dta
}
line coef q


-- Stas Kolenikov, PhD, PStat (SSC)
-- Senior Survey Statistician, Abt SRBI
-- Opinions stated in this email are mine only, and do not reflect the
position of my employer
-- http://stas.kolenikov.name



On Fri, Jun 14, 2013 at 9:28 AM, Kuziemko, Ilyana <ik2216@columbia.edu> wrote:
> Hi,
>
> I am trying to estimate the difference in the cost distributions of two
> different groups (group_a and group_b) for several different percentiles.
> The only complication is that I need to control for about 280 fixed
> effects, so I want the difference in the distributions netting out these
> effects.  I have about 1.4 million observations.
>
> Right now, I am estimating:
>
> forval x=.1(.1).9 {
> qreg cost group_a `list of fixed-effects' , quantile(`x')
> }
>
> Then I'll plot the _b[group_a] values for each value of `x'.
>
> Is there a way to speed up this calculation (which is currently taking
> forever)?  I just want to give a rough sense of where in the distribution
> the cost differences between these two groups are largest.  I am not going
> to use the standard errors at all and wonder if, perhaps, there was a way
> to speed up the regression in that case.
>
> Many thanks, ilyana
>
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/