{smcl} {hline} help for {hi:rsens} {hline} {title:Sensitivity analysis after matching with multiple nearest neighbours} {p 8 2 2}{cmdab:rsens} {it:outcomevar} {cmd:,} {cmdab:gamma}{cmd:(}{it:numlist}{cmd:)} {cmdab:nn}{cmd:(}{it:integer}{cmd:)} [{cmdab:treatment}{cmd:(}{it:varname}{cmd:)} {cmdab:id}{cmd:(}{it:varname}{cmd:)} {cmdab:matchid}{cmd:(}{it:namelist}{cmd:)} {cmdab:support}{cmd:(}{it:varname}{cmd:)}] {title:Description} {pstd} {cmd:rsens} calculates Rosenbaum sensitivity bounds following nearest neighbour matching with up to 20 nearest neighbours. While not mandatory, it is designed to be used after {help psmatch2}, as detailed below. {p_end} {pstd} {cmd:rsens} implements the procedure described by Rosenbaum (2002) to calculate bounds on the significance level for the treatment effect calculated using nearest neighbour matching, based on a signed rank statistic. {p_end} {title: Basic Syntax} (if used after {cmd:psmatch2}) {p 8 21 2}{cmdab:rsens} {it:outcomevar} {cmd:,} {cmdab:gamma}{cmd:(}{it:numlist}{cmd:)} {cmdab:nn}{cmd:(}{it:integer}{cmd:)} {phang} {it:outcomevar} is required, and specifies the variable name of the outcome for which treatment effects are being calculated. {p_end} {phang} {cmdab:gamma}{cmd:(}{it:numlist}{cmd:)} is required, and specifies the values of Gamma at which Rosenbaum's sensitivity bounds are to be calculated. {it: numlist} must contain 1, the level of Gamma at which allocation to treatment is assumed to be independent of unobservables. {p_end} {phang} {cmdab:nn}{cmd:(}{it:integer}{cmd:)} is required, and specifies the number of nearest neighbours with which matching was undertaken. {p_end} {title:Options} {pstd} {cmd:rsens} requires {it: outcomevar}, {cmdab:gamma}{cmd:(}{it:numlist}{cmd:)} and {cmdab:nn}{cmd:(}{it:integer}{cmd:)} to be specified if used after nearest neighbour matching undertaken with {cmd:psmatch2}. Otherwise, the following options might need to be specified. {p_end} {p 8 21 2}{cmdab:rsens} {it:outcomevar} {cmd:,} {cmdab:gamma}{cmd:(}{it:numlist}{cmd:)} {cmdab:nn}{cmd:(}{it:integer}{cmd:)} {cmdab:treatment}{cmd:(}{it:varname}{cmd:)} {cmdab:id}{cmd:(}{it:varname}{cmd:)} {cmdab:matchid}{cmd:(}{it:varname}{cmd:)} {cmdab:support}{cmd:(}{it:varname}{cmd:)} Where {phang} {opt treatment(varname)} provides the binary indicator for allocation to treatment. The default {it:varname} generated by {cmd:psmatch2} is {inp:_treated}. {phang} {opt id(varname)} provides the identifier used to specify matches. The default {it:varname} generated by {cmd:psmatch2} is {inp:_id}. {phang} {opt matchid(namelist)} specifies, for each treatment observation, the prefix for a list of variable names of the form var1, var2, var3... that identify matched control observations according to {cmd:id}{cmd:(}{it:varname}{cmd:)}. For instance, matching on the nearest 3 neighbours and specifying {cmd:matchid}{cmd:(}{it: mymatch}{cmd:)} implies that the dataset should contain three variables mymatch1, mymatch2 and mymatch3 that identify, for each treatment observation, the matched control observations using {cmdab:id}{cmd:(}{it:varname}{cmd:)} as the identifier. The default {it:varname} generated by {cmd:psmatch2} is {inp:_n}, as in _n1, _n2, _n3 for nearest 3 neighbour matching. {phang} {opt support(varname)} is an indicator variable that specifies the region of common support used in the matching exercise. The default {it:varname} generated by {cmd:psmatch2} is {inp:_support}. {pstd} The results of {cmd:rsens} are returned in r(rsensresult). {title:Example} {pstd}In the following, collgrad is the binary treatment and the outcome is wage. {inp: . sysuse nlsw88.dta, clear} {inp: . psmatch2 collgrad age race married, outcome(wage) n(2)} {inp: . rsens wage, gamma(1(0.5)4) nn(2)} {title:References} Rosenbaum, P.R. (2002) Observational Studies. 2nd edition. New York: Springer. {title:Author} {pstd} Sunil Mitra Kumar, King's College London; stuff.sunil@gmail.com