{smcl}
{* 29oct2003}{...}
{hline}
help for {hi:newey2} {right:(SSC distribution XX XXX 2002)}
{hline}

{title:Extended newey, regression with Newey-West standard errors}

{p 8 14}{cmd:newey2} {it:depvar} [{it:varlist1}]
          [{cmd:(}{it:varlist2}{cmd:=}{it:varlist_iv}{cmd:)}] [{it:weight}]
         [{cmd:if} {it:exp}] [{cmd:in} {it:range}]{cmd:,}
         {cmd:lag(}{it:#}{cmd:)} [{cmd:i(}{it:varname}{cmd:)} 
         {cmd:t(}{it:varname}{cmd:)} {cmd:force}
         {cmdab:nocon:stant} {cmdab:l:evel(}{it:#}{cmd:)} ]

{p 8}{cmd:neweyvif}

{p 8 14}{cmd:neweydmexog} [{it:varlist3}]


{title:Description}

{p}{cmd:newey2} provides extensions to Stata's official {cmd:newey}.  It accepts panel as well as time series data sets, and can instrument variables. For time series data without instruments, {cmd:newey2} behaves the same as {cmd:newey}.

{p}{cmd:neweyvif} works like Stata's official {cmd:vif} but can run after 
{cmd:newey}, as well as after {cmd:newey2} without instrumented variables. It 
computes variance inflation factors; see help {help vif}. {cmd:neweydmexog} is 
an extension of the unofficial Stata command {cmd:dmexog} that runs after 
{cmd:newey2} with instrumental variables. It tests for consistency of OLS after 
an instrumental variables regression; see {cmd:net search dmexog}.


{p}{cmd:aweight}s are allowed in {cmd:newey2}; see help {help weights}. They are not allowed in {cmd:neweydmexog}.

{p}{cmd:newey} shares the features of all estimation commands; see help
{help est}.


{p}The syntax of {help predict} following {cmd:newey2} is

{p 8 16}{cmd:predict} [{it:type}] {it:newvarname} [{cmd:if} {it:exp}]
         [{cmd:in} {it:range}] [{cmd:,} [ {cmd:xb} | {cmd:residuals} | {cmd:stdp} ] ]

{p}{cmd:xb}, {cmd:resid}, and {cmd:stdp} are available both in and out of sample; type
"{cmd:predict} {it:...} {cmd:if e(sample)} {it:...}" if wanted only for the
estimation sample.

{p}{cmd:newey2} produces Newey-West standard errors for coefficients estimated
by OLS or 2SLS.  The error structure is assumed to be heteroskedastic and
possibly autocorrelated up to some lag.

{p} For panel data sets, {cmd:newey2} computes pooled OLS or 2SLS estimates; it does not 
implement panel models such as fixed or random effects. {cmd: newey2} handles missing observations differently 
for time series and panel data sets. Consider the example of a time series data 
set containing gaps, which is then recast using {cmd:tsset} as a panel data set 
with one group. {cmd:newey} and {cmd:newey2} will not run on the time 
series version without {cmd:force}; with {cmd:force} they treat available 
observations as equally spaced.  After the set is cast as a panel, {cmd:newey2} 
will run without {cmd:, force}, and will assume zero serial correlation with missing observations.

{p} For both time series and panel data sets, {cmd:newey2} can instrument 
variables using the same syntax as {cmd: ivreg}; see help {help ivreg}. In 
computing the Newey-West sum for the second-stage regression, {cmd:newey2} uses 
residuals e=Y-XB where X contains all actual, not instrumented, values.

{p}If {cmd:lag(0)} is specified, the variance estimates produced by
{cmd:newey2} are the Huber/White/sandwich robust variance estimates 
calculated by {cmd:regress, robust} or {cmd:ivreg, robust}; see help {help regress} or help {help ivreg}.


{title:Options for newey2}

{p 0 4}{cmd:lag(}{it:#}{cmd:)} is not optional; it specifies the maximum lag
to consider in the autocorrelation structure.

{p 0 4}{cmd:t(}{it:varname}{cmd:)} specifies the variable recording the time
of each observation.  If this option is not used, t must already have been set using {cmd:tis} or {cmd:tsset}; see help {help tis} and help {help tsset}.  If the data set is a time series rather than a panel and observations are not equally spaced in time, you must use the {cmd:force} option.

{p 0 4}{cmd:i(}{it:varname}{cmd:)} specifies the variable recording the "group," 
"panel," or "independent unit" of each observation.  {cmd:newey2} will treat the 
data set as a panel set if this option is used, or if the set has already been 
configured as such with {cmd:iis} or {cmd:tsset}.

{p 0 4}For time series, {cmd:force} specifies that estimation is to be forced 
even though {cmd:t()} shows the data not to be equally spaced.  {cmd:newey2} 
will estimate the model assuming the lags based on the data ordered by {cmd:t()} 
are appropriate.  {cmd:force} has no effect for panel data sets.

{p 0 4}{cmd:noconstant} specifies that no intercept is to be included in the
model.

{p 0 4}{cmd:level(}{it:#}{cmd:)} specifies the confidence level, in percent,
for confidence intervals; see help {help level}.


{title:Options for {help predict}}

{p 0 4}{cmd:xb}, the default, calculates the linear prediction.

{p 0 4}{cmd:residuals} calculates the residuals of the linear prediction.

{p 0 4}{cmd:stdp} calculates the standard error of the linear prediction.


{title:Example}

{p 8 12}{inp:. newey2 usr idle sysv, lag(3) t(time)}{p_end}

{p 8 12}{inp:. newey2 usr idle (sysv = month temperature), lag(0)} {p_end}
{p 8 12}{ inp:. neweyvif} {p_end}

{p 8 12}{inp:. tsset countrycode period}{p_end}
{p 8 12}{inp:. newey2 gdpg lgdp (aid = egypt centam), lag(1)}{p_end}
{p 8 12}{ inp:. neweydmexog}{p_end}


{title:Author}

{p 8}David Roodman, Center for Global Development, USA{p_end}
{p 8}droodman@cgdev.org{p_end}

{title:Also see}

{p 1 14}Manual:  {hi:[U] 23 Estimation and post-estimation commands},{p_end}
{p 10 14}{hi:[U] 29 Overview of model estimation in Stata},{p_end}
	  {hi:[R] newey}
{p 10 14} {hi:[R] regress}, {hi:[R] regression diagnostics}{p_end}
{p 0 19}On-line:  help for {help est}, {help postest}; {help regress},
{help svyreg}, {help xtgls}, {help xtpcse}, {help regdiag}, {help regress}{p_end}