{smcl}
{* 15May2009}{...}
{hline}
help for {cmd:hodgesl}
{hline}
{cmd:Hodges-Lehmann aligned rank test}
{cmd:--------------------------------}
{cmd:Syntax}
{cmd:------}
{cmd:hodgesl} dataname varname blockname treatname savfile
{cmd:Description}
{cmd:-----------}
{cmd:hodgesl} implements Hodges-Lehmann aligned rank test of
the null hypothesis of no difference in an outcome variable
between treated and control observations over matched or
stratified sets. The test is an extension of the Wilcoxon
signed rank test to matching with multiple controls. It may
be employed in testing treatment effect that stratifies the
sample on single or multiple covariates, where the number of
strata compared to the number of total sample observations is
large, and within a stratum each treated subject has more than
one matched control. The test is needed when the analyst
evaluates average treatment effect and performs a significance
test of such effect after optimal matching.
{cmd:dataname} specifies the name of data file for the sample being
tested after matching or stratification. {cmd:varname} is the name
of outcome variable on which the analyst wants to test the
difference between treated and control observations. {cmd:blockname}
specifies the variable name that identifies matched or stratified
sets. {cmd:treatname} specifies the name of the dichotomous variable
identifying treatment conditions (i.e., treatname = 1 if the
observation receives treatment, and treatname = 0 if the
observation is a control). {cmd:savfile} specifies the name of a saved
data file for future analysis.
{cmd:Output and Return Values}
{cmd:------------------------}
After running {cmd:hodgesl}, Stata returns the sample average treatment effect
in the metric of the outcome variable {cmd:tx_effect}, Hodges-Lehmann mean
statistic {cmd:HL_mean}, Hodges-Lehmann standard-error statistic {cmd:HL_se},
the test statistic {cmd:z} that is the ratio of HL_mean and HL_se, and is
subject to a standard normal distribution, and the {cmd:p}-value of z via
which the analyst can perform a significance test of a nondirectional
hypothesis (i.e., perform a two-tailed test) or a directional hypothesis
(i.e. perform a one-tailed test).
After running {cmd:hodgesl}, {cmd:savfile} contains mean of the outcome and number of
observations for each treatment condition by matched set. The data file
can be used for postestimation analysis.
The analyst can use {cmd: return list} immediately after running
{cmd:hodgesl} to see statistics saved for further analysis.
{cmd:Examples}
{cmd:--------}
{cmd:. hodgesl cds lwss97 fm kuse fm_results}
{cmd:. hodgesl "C:\PSA\chapter5\cds.dta" lwss97 fm kuse "C:\tmp\fm_results"}
{cmd:. return list}
{cmd:. use "C:\tmp\fm_results", clear}
{cmd:. list}
{cmd:References}
{cmd:----------}
Guo, S., & Fraser, M. (2009). Propensity score analysis: Statistical
methods and applications. Thousand Oaks, CA: Sage Publications, Inc.
Lehmann, E. L. (2006). Nonparametrics: Statistical methods based on ranks
(Rev. ed., pp.132-141) New York: Springer.
Rosenbaum, P.R. (2002) Observational Studies. 2nd edition. New York: Springer.
{cmd:Author}
{cmd:------}
Shenyang Guo
University of North Carolina at Chapel Hill
sguo@email.unc.edu
{cmd:Also see:}
{cmd:---------}
{psee}Online: help for {helpb imbalance} if installed