stepwise.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}
{* 18mar2005}{...}
{cmd:help stepwise}{...}
{right:dialog:  {bf:{dialog stepwise}}}
{hline}

{title:Title}

{p2colset 5 21 23 2}{...}
{p2col:{hi:[R] stepwise} {hline 2}}Stepwise estimation{p_end}
{p2colreset}{...}


{title:Syntax}

{p 8 18 2}
{cmd:stepwise}
	[{cmd:,}
		{it:options}
	] {cmd::} {it:command}

{synoptset 16 tabbed}{...}
{synopthdr}
{synoptline}
{syntab:Model}
{p2coldent:* {opt pr(#)}}significance
	level for removal from the model{p_end}
{p2coldent:* {opt pe(#)}}significance
	level for addition to the model{p_end}

{syntab:Model2}
{synopt:{opt forw:ard}}perform forward-stepwise selection{p_end}
{synopt:{opt hier:archical}}perform hierarchical selection{p_end}
{synopt:{opt loc:kterm1}}keep the first term{p_end}
{synopt:{opt lr}}perform likelihood ratio-test instead of Wald test{p_end}
{synoptline}
{p2colreset}{...}
{p 4 6 2}
* At least one of {opt pr(#)} or {opt pe(#)} must be specified.{p_end}
{p 4 6 2}
{cmd:by} and {cmd:xi} may be used with {cmd:stepwise}; see {help prefix}.{p_end}
{p 4 6 2}
Weights are allowed if {it:command} allows them; see {help weight}.{p_end}
{p 4 6 2}
All postestimation commands behave as they would after {it:command} without
the {opt stepwise} prefix; see the postestimation manual entry for
{it:command}.
{p_end}


{title:Description}

{pstd}
{opt stepwise} performs stepwise estimation.  Typing

{pin}
{cmd:. stepwise,} {opt pr(#)}{cmd::} {it:command}

{pstd}
performs backward-selection estimation for {it:command}.  The stepwise
selection method is determined by the following option combinations:

{synoptset 24}{...}
{synopthdr}
{synoptline}
{synopt:{opt pr(#)}}backward selection{p_end}
{synopt:{opt pr(#)} {opt hierarchical}}backward hierarchical selection{p_end}
{synopt:{opt pr(#)} {opt pe(#)}}backward stepwise{p_end}

{synopt:{opt pe(#)}}forward selection{p_end}
{synopt:{opt pe(#)} {opt hierarchical}}forward hierarchical selection{p_end}
{synopt:{opt pr(#)} {opt pe(#)} {opt forward}}forward stepwise{p_end}
{synoptline}
{p2colreset}{...}

{pstd}
{it:command} defines the estimation command to be executed.  The following
Stata commands are supported by {cmd:stepwise}.

{pmore}
{helpb clogit}{space 2}
{helpb cloglog}{space 2}
{helpb cnreg}{space 2}
{helpb glm}{space 2}
{helpb intreg}{space 2}
{helpb logistic}{space 2}
{helpb logit}{space 2}
{helpb nbreg}{space 2}
{helpb ologit}{space 2}
{helpb oprobit}{space 2}
{helpb poisson}{space 2}
{helpb probit}{space 2}
{helpb qreg}{space 2}
{helpb regress}{space 2}
{helpb scobit}{space 2}
{helpb stcox}{space 2}
{helpb streg}{space 2}
{helpb tobit}

{pstd}
{cmd:stepwise} expects {it:command} to have the following form:

{p 8 21 2}
{it:command_name} [{depvar}] {it:term} [{it:term} ...] {ifin} {weight}
[{cmd:,} {it:command_options}]

{pstd}
where {it:term} is either {varname} or {opth (varlist)} (a
{it:varlist} in parentheses indicates that this group of variables is to be
included or excluded together).  Note that {it:depvar} is not present when
{it:command_name} is {helpb streg} or {helpb stcox}; otherwise, {it:depvar} is
assumed to be present.  For {helpb intreg}, {it:depvar} is actually two
dependent variable names ({it:depvar1} and {it:depvar2}).

{pstd}
{cmd:sw} is a synonym for {cmd:stepwise}.


{title:Options}

{dlgtab:Model}

{phang}
{opt pr(#)} specifies the significance level for removal
from the model; terms with p>={opt pr()} are eligible for removal.

{phang}
{opt pe(#)} specifies the significance level for addition to
the model; terms with p<{opt pe()} are eligible for addition.

{dlgtab:Model2}

{phang}
{opt forward} specifies the forward-stepwise method and may only be specified
when both {opt pr()}
and {opt pe()} are also specified.  Specifying both {opt pr()} and {opt pe()}
without {opt forward} results in backward-stepwise selection.
Note that specifying only
{opt pr()} results in backward selection, and specifying only {opt pe()}
results in forward selection.

{phang}
{opt hierarchical} specifies hierarchical selection.

{phang}
{opt lockterm1} specifies that the first term be included in the
model and not be subjected to the selection criteria.

{phang}
{opt lr} specifies the test of term significance be the
likelihood-ratio test.  The default is the less computationally expensive Wald
test; that is, the test is based on the estimated variance{c -}covariance
matrix of the estimators.


{title:Examples}

{pstd}Backward selection{p_end}
{phang2}
{cmd:. stepwise, pr(.2): reg price mpg weight displ}
{p_end}
{phang2}
{cmd:. stepwise, pr(.2): reg price mpg weight displ (r1-r4)}
{p_end}

{pstd}Backward hierarchical selection{p_end}
{phang2}
{cmd:. stepwise, pr(.2) hier: reg price mpg weight displ (r1-r4)}
{p_end}
{phang2}
{cmd:. stepwise, pr(.2) hier: reg price mpg (weight displ) (r1-r4)}
{p_end}

{pstd}Forward selection{p_end}
{phang2}
{cmd:. stepwise, pe(.2): logit outcome (sex weight) treated1 treated2}
{p_end}
{phang2}
{cmd:. stepwise, pe(.2) lock: logit outcome (sex weight) treated1 treated2}
{p_end}

{pstd}Forward hierarchical selection{p_end}
{phang2}
{cmd:. stepwise, pe(.2) hier: logit outcome (sex weight) treated1 treated2}
{p_end}
{phang2}
{cmd:. stepwise, pe(.2) lock hier: logit outcome (sex weight) treated1 treated2}
{p_end}

{pstd}
Reporting saved results{p_end}
{phang2}
{cmd:. stepwise}


{title:Programming for stepwise}

{pstd}
{cmd:stepwise} requires that {it:command_name} follow
{help language:standard Stata syntax} and allow the {helpb if} qualifier.
Furthermore, {it:command_name} must have {opt sw} or {opt swml} as a program
property; see {helpb program}.  If {it:command_name} has {opt swml}
as a property, {it:command_name} must save the log-likelihood value in
{cmd:e(ll)} and model degrees of freedom in {cmd:e(df_m)}.


{title:Also see}

{psee}
Manual:  {bf:[R] stepwise}

{psee}
Online:  
{helpb clogit},
{helpb cloglog},
{helpb cnreg},
{helpb glm},
{helpb intreg},
{helpb logistic},
{helpb logit},
{helpb nbreg},
{helpb ologit},
{helpb oprobit},
{helpb poisson},
{helpb probit},
{helpb qreg},
{helpb regress},
{helpb scobit},
{helpb stcox},
{helpb streg},
{helpb tobit}
{p_end}