.-
help for ^rbounds^
.-
.
^Rosenbaum sensitivity analysis for average treatment effects on the treated^
^---------------------------------------------------------------------------^
.
^rbounds^ varname [^if^ exp]^,^ ^g^amma^(^numlist^)^ [^a^lpha^(^#^)^ ^acc(^#^)^ ^sigonly^ ^dots^]
.
.
^Description^
^-----------^
.
^rbounds^ calculates Rosenbaum bounds for average treatment effects on
the treated in the presence of unobserved heterogeneity (hidden bias)
between treatment and control cases. ^rbounds^ takes the difference in the
response variable between treatment and control cases as input variable
^varname^. The procedure then calculates Wilcoxon signrank tests that give
upper and lower bound estimates of significance levels at given levels of
hidden bias. Under the assumption of additive treatment effects, rbounds
also provides Hodges-Lehmann point estimates and confidence intervals for
the average treatment effect on the treated. If installed, the input variable
^varname^ may be generated from @psmatch@ or @psmatch2@. Currently,
^rbounds^ implements the sensitivity tests for matched (1x1) pairs only.
.
.
^Options^
^-------^
.
^gamma(^numlist^)^ specifies the values of cap gamma for which to carry out
the sensitivity analysis. Estimates at cap gamma = 1 (no heterogeneity)
are included in the calculations by default. ^gamma()^ is required by
rbounds.
.
^alpha(^#^)^ specifies the values of alpha in the calculation of confidence
intervals for the Hodges-Lehmann point estimate of the average treatment
effect.
.
^acc(^#^)^ specifies the convergence criterion of the line search algorithm
used to find the Hodges-Lehmann point estimates. Convergence level is set
to 1e-acc, the preset value is acc=6.
.
^sigonly^ restricts ^rbounds^ to calculate Wilcoxon signrank tests for
significance levels only.
^dots^ may be specified for status information. The option is useful for
checking total execution time with large samples.
.
.
^Examples^
^--------^
.
^(1) post-estimation use after psmatch2 (or related) command^
.
. psmatch2 college, outc(wage) pscore(pscore) caliper(.25) common
.
. gen delta = wage - _wage if _treat==1 & _support==1
.
. rbounds delta, gamma(1 (1) 3) - performs sensitivity analysis
at gamma = 1,2,3.
.
. rbounds delta, gamma(1 (1) 3) alpha(.90) - as above, but uses a=.90 for
confidence interval calculations.
.
. rbounds delta, gamma(1 (1) 3) sigonly - performs sensitivity analysis
at gamma = 1,2,3 without calcula-
ting the Hodges-Lehmann estimates.
.
^(2) using Rosenbaum (2002, p.82) data^
.
. use rbtest.dta, clear
.
. rbounds dlead, gamma(1 (1) 4 4.25 5 6) - replicates analyses of Chapter 4.3
in Rosenbaum (2002)
.
.
.
^References^
^----------^
.
Rosenbaum, P.R. (2002) Observational Studies. 2nd edition. New York: Springer.
.
DiPrete, T.A., Gangl, M. (2004) Assessing Bias in the Estimation of Causal Effects:
Rosenbaum Bounds on Matching Estimators and Instrumental Variables Estimation
with Imperfect Instruments. Discussion paper SP I 2004-101. Berlin: WZB.
.
.
^Author^
^-------^
.
Markus Gangl
Social Science Centre Berlin (WZB)
gangl@wz-berlin.de
.
.
^Also see:^
^---------^
.
On-line: help for @psmatch@, @psmatch2@ if installed