vce_option.hlp

是一个经济学管理应用软件 很难找的 但是经济学学生又必须用到

{smcl}
{* 29mar2005}{...}
{cmd:help vce_option}{right:also see:  {help bootstrap postestimation}}
{right:after {cmd:vce(bootstrap)}}
{hline}

{title:title}

{p2colset 5 23 25 2}{...}
{p2col:{bf:[R] {it:vce_option}} {hline 2}}Alternative variance estimators
{p_end}
{p2colreset}{...}


{title:Syntax}

{phang}
{opt vce(vcetype)} specifies how to perform variance-covariance estimation.
Allowed are

{p 8 12 2}
{cmd:vce(oim)}

{p 8 12 2}
{cmd:vce(}{cmdab:r:obust)}

{p 8 12 2}
{cmd:vce(opg)}

{p 8 12 2}
{cmd:vce(}{cmdab:boot:strap} [{cmd:,}
        {it:{help bootstrap:bootstrap_options}}]{cmd:)}

{p 8 12 2}
{cmd:vce(}{cmdab:jack:knife} [{cmd:,}
        {it:{help jackknife:jackknife_options}}]{cmd:)}


{title:Description}

{pstd}
This entry describes the {opt vce()} option, which is common to most
estimation commands.  {cmd:vce()} specifies how the variance-covariance
matrix of the estimators is to be estimated.


{title:Options}

{dlgtab:SE/Robust}

{phang}
{cmd:vce(oim)} is usually the default if {cmd:vcetype()} is not specified.
{cmd:oim} stands for observed information matrix, which is the inverse of the
negative matrix of second derivatives; see {helpb ml}.

{phang}
{cmd:vce(robust)} uses the robust or sandwich estimator of variance.
Estimation commands that allow option {cmd:vce(robust)} also allow
option {cmd:robust} as a synonym; it does not matter which you specify.
Many estimation commands that allow {cmd:vce(robust)} also allow a 
{opt cluster(varname)} option; see
{help estimation options##robust:estimation options}.

{phang}
{cmd:vce(opg)} uses the sum of the outer product of the gradient vectors; 
see {helpb ml}.

{phang}
{cmd:vce(bootstrap} [{cmd:,} {it:bootstrap_options}]{cmd:)} uses a 
nonparametric bootstrap; see {helpb bootstrap}.  After estimation with
{cmd:vce(bootstrap)}, see {helpb bootstrap postestimation} to obtain
percentile-based or bias-corrected confidence intervals.

{phang}
{cmd:vce(jackknife} [{cmd:,} {it:jackknife_options}]{cmd:)} uses the
delete-one jackknife; see {helpb jackknife}.


{title:Remarks}

{pstd}
Remarks are presented under the headings

{pmore}
{helpb vce_option##Prefix_commands:Prefix commands}{break}
{helpb vce_option##Passing_options:Passing options in vce()}


{marker Prefix_commands}{...}
{title:Prefix commands}

{pstd}
In many cases, specifying {cmd:vce(bootstrap)} or {cmd:vce(jackknife)} is
equivalent to using the respective prefix command.
Here is an example using {opt jackknife} with {helpb regress}.

	{cmd:. webuse auto}
        {cmd:. regress mpg turn trunk, vce(jackknife)}
        {cmd:. jackknife: regress mpg turn trunk}

{pstd}
However, you should use the {opt vce()} option whenever available
because the equivalence of the prefix commands is not true for all estimation
commands.  For example, to properly use the {opt jackknife} prefix
command with {helpb clogit} (without the {opt cluster()} option), you must
inform {opt jackknife} to leave out observations from each group identified in
the {opt group()} option of {opt clogit}.

	{cmd:. webuse clogitid}
        {cmd:. jackknife, cluster(id): clogit y x1 x2, group(id)}

{pstd}
This extra information is automatically communicated to {cmd:jackknife} by
{cmd:clogit} when the {opt vce()} option is specified.

        {cmd:. clogit y x1 x2, group(id) vce(jackknife)}


{marker Passing_options}{...}
{title:Passing options in vce()}

{pstd}
As indicated above, specifying the {cmd:vce(bootstrap)} or
{cmd:vce(jackknife)} option calls a respective prefix command.
You can pass options directly to the prefix command by supplying them
within the {opt vce()} option.  Here is an example in which 300
bootstrap replications are requested and saved in {cmd:bsreg.dta}.

        {cmd:. webuse auto, clear}
        {cmd:. regress mpg turn trunk, vce(bootstrap, rep(300) saving(bsreg))}
	{cmd:. bstat using bsreg}


{title:Also see}

{psee}
Manual:  {bf:[R] {it:vce_option}}

{psee}
Online:  {helpb bootstrap},
{helpb jackknife},
{helpb ml}
{p_end}