{smcl}
{* Version 1, september 2006}{...}
{hline}
help for {hi:checkrob} {center: (Mikkel Barslund, September 2006)}
{hline}
{title:Test of robustness of specification}
{p 4 12 2}{cmd:checkrob} {it:#corevars} {it:#testingvars} [options] [filename.txt] [noisily] : {it:estimation command}
{p 4} {it:where}
{p 6 16 2} {hi:#corevars} is the number of core variables (to be retained in all regressions).
{p 6 19 2} {hi:#testingvars} is the number of testing variables (to be systematically included/excluded).
{p 6 16 2} {hi:options} are additional output options. All return variables [e(name)] can be specified with a space between them. Default (=none).
{p 6 21 2} {hi:filename.txt} is an option for specifying the name of the output file. Note, that it must have the ".txt" extension.
Default (={it:result.txt}). The table file will get the name table_{it:filename.txt}. If {it:filename.txt}
already exists {cmd:checkrob} will end with an error message.
{p 6 14 2} {hi:noisily} displays the output for each regression. Default (=quietly).
{p 6 25 2} {it}{hi:estimation command}{sf} can be any single equation estimation command and
some 2-equation estimation commands where the second equation is fixed (such as
{help heckman:heckman} or {help heckprob:heckprob}
with a fixed (unaltered) selection equation in all regressions).{break}
{it}{hi:Note:}{sf} the order of the explanatory variables is important for {cmd:checkrob} to work
properly. First should the core variables be specified and only thereafter the testing
variables (see {hi:Examples}{sf} below).{break}
A full set of options supported by the estimation command can be applied to each estimation (e.g. {help weights:weights},
{help cluster:cluster}, {help in:in range}, {help if:if} etc.)
{p 6 25 2} {hi:checkrob} supports survey settings via the {help svy:svy:} prefix (see also help {help svyset:svyset}).
{p 6 16 2} {it}{hi:OBS...OBS}{sf} {cmd:checkrob} only does limited syntax checking. Thus, the user should check
the prerun explanation by {cmd:checkrob} of what it is about to do carefully.
{title:Description}
{p 4 4 2}{cmd:checkrob} estimates a set of regressions where the dependent variable is regressed (with whatever method
is specified in {it:estimation command}) on the {it:core} variables {c -} which are included in all regressions {c -}
and all possible
combinations of the {it:testing variables}. In total 2^{it:#testingvars} regressions are generated. For each
regression the coefficients, their standard errors and optional estimation return variables
({help ereturn:ereturn list}) are recorded in comma separated format in the file named {it:filename.txt}. This file
can be accessed after {cmd:checkrob} has run with a standard text editor. Further, {cmd:checkrob} produces a second text
file table_{it:filename.txt} including the following statistics for each {it:core} and {it:testing} variable:
maximum (Max), minimum (Min) and average value (Mean) of the coefficient over all regressions, average standard
deviation (AvgSTD), share of regressions where the coefficient is significant (PercSigni), share of regressions
where the sign is positive (Perc+), respective negative (Perc-), the average t-value (AvgT) and the number of times
the variable was included in the regressions (Obs) (this might differ between core variables (and between testing variables)
if a variable is sometimes dropped. For each variable the average over optional estimation return variables in
these regression where the variable featured is also reported.
{p 4 4 2}The table will look like this (recorded in comma separated form):
{center: table_{it:filename.txt} }
{center: Core var Max Min Mean AvgSTD PercSigni Perc+ Perc- AvgT Obs e(r2)}
{center:{hline 86}}
{center: Education 4 }
{center: Experience 4 }
{center: age 4 }
{center: Testing Max Min Mean AvgSTD PercSigni Perc+ Perc- AvgT Obs e(r2)}
{center:{hline 86}}
{center: Educ.spouse 2 }
{center: Children 2 }
{p 4 4 2}{cmd:checkrob} is a useful tool if you want to check the robustness of including/excluding one
or more controls in your preferred specification. {cmd:checkrob} was used to produce the robustness results
in Barslund et al. (2007). See the command {help rcheck:rcheck} (if installed) by Kenneth L. Simons for a similar procedure,
however {help rcheck:rcheck} cannot handle multiple equation such as the {help heckman:heckman} estimation command.
{title:Examples}
{p 8 12 2}{cmd:. checkrob} {it:#corevars} {it:#testingvars} [options] [filename.txt] [noisily] : {it:estimation command}
{p 8 10 2}{cmd:. checkrob 2 3: reg wage education experience age educspouse children}{break}
Here education and experience are core variables; age, educspouse and children are testing variables.
{p 8 10 2}{cmd:. checkrob 1 4: reg wage education experience age educspouse children}{break}
Here education is the only core variables; experience, age, educspouse and children are testing variables.
{p 8 10 2}{cmd:. checkrob 1 4: reg wage education experience age educspouse children, cluster(ea)}
{p 8 10 2}{cmd:. checkrob 2 2 output.txt noisily: reg wage education experience age children if children>4, cluster(ea)}
{title:Author}
{p 4 4 2} Please address comments and suggestions to:
{p 4 4 2} Mikkel Barslund, Department of Economics, University of Copenhagen, Denmark {break}
{title:References}
{p 4 4 2} Barslund, M., Chiconela, J., Rand, J. & Tarp, F., 2007,{break}
'Understanding Victimization: The Case of Mozambique'. World Development Vol. 35, No. 7 (July)
{title:Also see}
{p 4 19 2}Online: help for {help rcheck} if installed{p_end}