Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down at the end of May, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Quantile diff-in-diff and standard errors


From   Prashant <presearchwork@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   st: Quantile diff-in-diff and standard errors
Date   Wed, 29 Dec 2010 17:47:22 +0800

Dear all,

I am trying to use quantile difference-in-differences (where my
regression equation also has a variables for year and an interactions
with year to capture the effects of time trends). Specifically, there
was an intervention that applied to 9 out of 23 counties/clusters at
year 6 (and onwards) of the ten years of data I have available. I am
looking at student-level exam score outcomes for about 300,000
students, with essentially no missing data.

The basic model is:

score = b0 + b1*year + b2*treatment_area + b3*post_policy +
b4*Treatment_area*year + b5*post_policy*year + b6*
Treatment_area*post_policy + b7*(Treatment_area*post_policy*year) +
error

I ran the following code (based on helpful previous posts, see below
for a list):

capture program drop bootqreg
prog bootqreg
qreg score year Tr-area post-policy Tr-area_year post-policy_year
Tr-area_post-policy Tr-area_post-policy_year, quantile(.5)
end

bs, reps(200) cluster(county_hk): bootqreg

When I bootstrap the quantile regression to account for clustering at
the county level (county_hk), I get "red Xs" about 1/3 of the time
(and the associated message "one or more parameters could not be
estimated in @@ bootstrap replicates; standard-error estimates include
only complete replications.")

My questions are:
1) Am I getting red Xs during the bootstrap because some bootstrap
samples have no treatment counties (i.e. counties in the treatment
area) or no control counties? Or is there some other reason?
2) Are there any other steps I can take to estimate the standard
errors correctly?

I may just be making some fundamental mistake somewhere...any help
would be most appreciated.

I have read a number of helpful posts about clustering errors with
quantile regression:
http://www.stata.com/statalist/archive/2009-11/msg01052.html
http://www.stata.com/statalist/archive/2010-03/msg01306.html
http://www.stata.com/statalist/archive/2007-12/msg00162.html
http://www.stata.com/statalist/archive/2007-09/msg00147.html


For reference, Quantile Difference-in-Differences has been discussed
by (among others):
 Athey S and Imbens G (2006), "Identification and inference on
nonlinear difference-in-differences models," Econometrica, 74(2),
431–497.
 Meyer, B., K. Viscusi and D. Durbin (1995), “Workers Compensation and
Injury Duration: Evidence from a Natural Experiment,” American
Economic Review, 85(3), 322-340.

Thank you very much,
Prashant

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index