{smcl}
{* version 1.1 - Jan., 20, 2007}{...}

{cmd:help grubbs} 
{hline}

{title:Title}

{p2colset 5 16 21 2}{...}
{p2col:{hi: grubbs} {hline 2}} Performs Grubbs' test for outliers

{p2colreset}{...}


{title:Syntax}

{p 8 15 2}
{cmd:grubbs}
{varlist}
{ifin}
[, {it:options}]

{synoptset 22 tabbed}{...}
{synopthdr}
{synoptline}
{syntab:Procedure}
{synopt:{opt it:er(#)}} perform maximum of # iterations; default is {opt it:er(16000)}{p_end}
{synopt:{opt le:vel(#)}} set confidence level; default is {opt le:vel(95)} {p_end}

{syntab:Output}
{synopt:{opt dr:op}}  drop out of the sample observations that are outliers {p_end}
{synopt:{opt gen:erate(newvar1 ...)}} Variable(s) to be created, containing 1 if the observation is an outlier{p_end}

{syntab:Reporting}
{synopt:{opt lo:g}} display an iteration log; default is no log.  {p_end}
{synoptline}
{p2colreset}{...}
{p 4 6 2}

{title:Description}

{p 4 4 2}{cmd:grubbs} test (Grubbs, 1969 and Stefansky, 1972) can be used to detect outliers in a data set, either 
creating a new variable (equal to 1 if the observation is an outlier and 0 otherwise) or dropping outliers out 
of the data set. The Grubbs test is also known as the maximum normed residual test. The Grubbs test detects one 
outlier at each iteration. The outlier is expunged from the data set and the test is iterated until no outliers 
remain.

{title:Options}

{dlgtab:Procedure}

{phang}
{opt it:er(#)} set maximum number of iterations; default is {opt it:er(16000)}.

{phang}
{opt le:vel(#)} set confidence level; default is {opt le:vel(95)}.


{dlgtab:Output}

{phang}
{opt dr:op} drop out of the sample the observations that are outliers in the Grubbs sense (not to be used in conjunction with {opt gen:erate}). 

{phang}
{opt gen:erate(newvar1 newvar2 ...)} create new Variables {opt (newvar1 newvar2 ...)}, 
        containing 1 if the observation is an outlier in the Grubbs sense and 0 otherwise. If no 
        {opt gen:erate} option is specified or if the number of new variable names does not match, default variable names will be used. 

{dlgtab:Reporting}

{phang}
{opt lo:g} display an iteration log; default is no log.

{phang}


{title:Examples}

{phang}{cmd:. sysuse auto.dta}{p_end}

{phang}{cmd:. grubbs price mpg}{p_end}

{phang}{cmd:. grubbs price mpg, gen(gr_pr gr_mpg) log drop iter(1000) level(90)}{p_end}

{phang}{cmd:. grubbs price mpg, log gen(gru_pr gru_mpg)}{p_end}
{phang}{cmd:. list if gru_pr}{p_end}
{phang}{cmd:. reg  price mpg if ~gru_pr & ~gru_mpg}{p_end}


{title:Author} 

        Nicolas Couderc, 
	Reims Management School and Université Paris 1 (Panthéon - Sorbonne),
        couderc@univ-paris1.fr

        Questions, comments and bug reports are welcome.


{title:References} 

Grubbs F. (1969), Procedures for Detecting Outlying Observations in Samples, {it:Technometrics}, 11(1), 1-21.
Stefansky W. (1972), Rejecting Outliers in Factorial Designs, {it:Technometrics}, 14, 469-479.

{title:Acknowledgements} 

{p 4 4 2}This program owes a debt to Anne-Célia Disdier and Keith Head for the original Grubbs procedure implementation in Stata (available at http://strategy.sauder.ubc.ca/head/grubbs.ado) and to {cmd:hadimvo}.

{title:Also see}

{p 4 13 2}On-line:  help for {help hadimvo}.