Title
next -- Regression discontinuity (RD) estimator
Syntax
next outcomevar assignmentvar [if] [in] , [options]
options Description
---------------------------------------------------------------------------
Model
regtype(string) regression type
threshold(real) threshold
data_min(int) minimum number of data points that must be used to
predict the next value
p1(int) minimum order of the polynomial
p2(int) maximum order of the polynomial
base(real) base weight for the weighting scheme
mspe_min(int) minimum number of MSPEs that are allowed to be
included in a weighted average of MSPE
confidence(real) confidence interval
bin_left(int) number of bins used set by the user
bin_right(int) number of bins used set by the user
Reporting
details print details
---------------------------------------------------------------------------
Description
next simultaneously selects the polynomial specification and bandwidth that
minimizes the predicted mean squared error at the threshold of a
discontinuity. It achieves this selection by evaluating the
combinations of specification and bandwidth that perform best in
estimating the next point in the oberved sequence on each side of the
discontinuity. For more information on the model, see Next: An Improved
Method for Identifying Impacts in Regression Discontinuity Design.
Options
+-------+
----+ Model +--------------------------------------------------------------
regtype(string) sets the regression type. Values can be regress, probit, or
logit. If not specified by user, regression type is regress.
threshold(real) sets the threshold of the discontinuity. If not specified
by the user, the threshold is assumed to be 0. Note that if X=T, the
observation is assumed to be to the right of the threshold.
data_min(int) sets the minimum number of data points that must be used to
predict the next value in the series. Note that value should be in the
range of 1 to min(number of distinct values of X on left of
threshold-1, number of distinct values of X on right of threshold-1).
If data_min>1, then random walk specification is skipped. If the user
does not specify, the minimum number of data points is set at 5.
p1(int), p2(int) sets the minimum(maximum) order of the polynomial to be
tested. If the user does not specify, the minimum is an order of 0 and
the maximum is an order of 5.
base(real) sets the base weight for the weighting scheme and should lie in
the interval of 1 (uniform weight) to infinity. If the user does not
specify, the base weight is set at 1000.
mspe_min(int) sets the minimum number of MSPEs that are allowed to be
included in a weighted average of MSPE. Must be >=2. If the user does
not specify, the minimum number of MSPEs is 5.
confidence(real) sets the confidence interval and can be between 0 and 100.
If the user does not specify, the confidence interval is set at 80.
bin_left(int), bin_right(int) sets the number of bins used on each side of
the discontinuity. If set by the user, then bin_left and bin_right is
set equal to Min(user set bin size, Min(matsize,maxvar)/(number of
specifications used)-2) on each side of the discontinuity. If not set
by the user, then the bins on each side of discontinuity is set equal
to Min(100, number of distinct values on that side of the
discontinuity).
+-----------+
----+ Reporting +----------------------------------------------------------
details prints the details of each sepcification tested by the model.
Examples
Setup (data from Jacob et al., 2012, with simulated treatment effect of -10
at threshold=215)
import excel using http://www.mdrc.org/sites/default/files/img/RDD_Guide
> _Dataset_0.xls, firstrow sheet("Data")
Run regression discontinuity analysis with default settings, except
threshold set at 215 (as in Jacob et al., 2012)
next posttest pretest, t(215)
Author
Mark C. Long, University of Washington (corresponding author)
Email marklong@uw.edu
Jordan Rooklyn, University of Washington
References
Long, M., and J. Rooklyn. 2016. Next: An Improved Method for Identifying
Impacts in Regression Discontinuity Design.. Working paper.
Jacob, R.T., P. Zhu, M.A. Somers, and H. Bloom. 2012. A Practical Guide to
Regression Discontinuity. New York: MDRC.
See Also
rd (if installed) ssc install rd (to install this command)
(remote file ends)
---------------------------------------------------------------------------------
(click here to return to the previous screen)