asmprobit.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

{space 8}{c BLC}{space 12}{c BRC}{space 13}{c BLC}{space 12}{c BRC}{break}

{pmore}{cmd:correlation(fixed} {it:matname}{cmd:)} identifies fixed and free correlation 
parameters.  The fixed parameters are identified by numeric values between -1 and 1
and free parameters (those that are to be estimated) are identified by missing values (.).

{pmore}Only the lower triangular part of the matrix is accessed and the order of the
rows and columns must match the numeric order of the alternative levels.

{phang}{opt stddev(stddev)} specifies the variance structure of the latent-variable 
errors.  {opt stddev(heteroskedastic)},
the default, has J-2 estimable parameters.  The error standard deviations
for the alternatives specified in {opt basealternative()} and 
{opt scalealternative()} are
fixed to one.  {opt stddev(homoskedastic)} constrains all the standard deviations
to equal one.

{pmore}Advanced users can use {cmd:stddev(pattern} {it:matname}{cmd:)} or 
{cmd:stddev(fixed} {it:matname}{cmd:)} to 
supply the name of a 1 by J matrix identifying the 
the variance structure of their choosing.  The {it:matname} for the {opt pattern}
option contains sequential positive integers,
starting from 1, that identify each standard deviation parameter to be estimated.  The integers
can be repeated to identify alternatives that have latent-variable errors with equal 
standard deviation.
For example, assume J = 4 with alternative levels 1, 2, 3, 4, and it is natural
for the latent-variables associated with alternatives 1 and 2 to have equal error variances 
while those associated with alternatives 3 and 4 
should have equal variances.  Using {cmd:basealternative(}{it:1}{cmd:)} and 
{cmd:scalealternative(}{it:2}{cmd:)} with the following matrix will parameterize this
model: 

{pmore}{space 10}1{space 2}2{space 2}3{space 2}4{break}
{space 8}{c TLC}{space 12}{c TRC}{break} 
{space 8}{c | } {bf:.}{space 2}{bf:.}{space 2}1{space 2}1{space 1}{c |}{break}
{space 8}{c BLC}{space 12}{c BRC}{break}

{pmore}It may be more convenient instead to use the {opt fixed} option 
with a 1 x J matrix {it:matname} to identify the free and
fixed standard deviations.  The free standard deviations (those that are to be 
estimated) are identified by missing values (.), while the fixed standard deviations are
identified by positive numeric values.  The order of the matrix entries must
be the same as the numeric order of the alternatives.

{phang}{opt noconstant} suppresses the J-1 alternative-specific constant terms.

{phang}{opt basealternative(#)} specifies the alternative used to normalize the
latent-variable location (also referred to as the level of utility).  The
standard deviation for the latent-variable error associated with the base
alternative is fixed to one and its correlations with all other
latent-variables' errors are set to zero.  The default is the first
alternative.  Note, however, that if a {it:fixed} or {it:pattern} matrix is
given in the {opt stddev()} and the {opt correlation()} options, then the 
{opt basealternative()} will be implied by the fixed standard deviations and
correlations in the matrix specifications.  {opt basealternative()} cannot be
equal to {opt scalealternative()}.

{phang}{opt scalealternative(#)} specifies the alternative used to normalize
the latent-variable scale (also referred scale of utility).  The default is to
use the second alternative.  Note, however, that if a {it:fixed} or
{it:pattern} matrix is given in the {opt stddev()} option, then the 
{opt scalealternative()} will be implied by the fixed standard deviations in the
matrix specification.  {opt scalealternative()} cannot be equal to the 
{opt basealternative()}.

{dlgtab:SE/Robust}

{phang}{opt vce(vcetype)}; see {it:{help vce_option}}. 

{phang}{opt robust}, {opth cluster(varname)}; see {help estimation options}.

{dlgtab:Reporting}

{phang}{opt notransform} prevents transforming the Cholesky factored variance-covariance
estimates to their correlations and standard deviations representations.  If 
{opt correlation(unstructured)} and {opt stddev(heteroskedastic)} are specified, the
variance-covariance of the errors are parameterized using a Cholesky factorization.
Recovering the standard deviations and correlations requires an additional optimization step,
and for large datasets this reparameterization can be slow.

{pmore} Enforcing a specific variance-covariance structure is not generally possible under
the Cholesky factored parameterization, so {cmd: asmprobit} uses the standard deviations
and correlations during optimization if either {opt correlation(unstructured)} or 
{opt variance(heteroskedastic)} is not used.  In these cases {opt notransform} has no
effect.{p_end}

{phang}{opt level(#)}; see {help estimation options}.

{dlgtab:Int options}

{phang}{opt intmethod(hammersley|halton|random)} specifies the
method of generating the points used in the pseudo- or quasi-Monte Carlo integration of
the multivariate normal density.  Specify {opt intmethod(hammersley)}, the default, 
to use the Hammersley 
sequence, {opt intmethod(halton)} to use the Halton sequence, and 
{opt intmethod(random)} to use a sequence of uniform random numbers.

{phang}{opt intpoints(#)} specifies the number of points to use in the 
pseudo- or quasi-Monte Carlo integration.  If this option is not specified, the number of
points is 50 x J if {opt intmethod(hammersley)} or {opt intmethod(halton)} is used
and 100 x J if {opt intmethod(random)} is used.  Larger values of {opt intpoints} 
provide better approximations to the log likelihood, but at the cost of added computational time.

{phang}{opt intburn(#)} specifies where in the Halton Hammersley or Halton sequence
to start.  This helps reduce the correlation between the sequences of each
dimension.  The default is 0.  This option is ignored if {opt intmethod(random)} 
is specified.

{phang}{opt intseed(code)} specifies the seed to use for generation the uniform
sequence.  This option only applies for {opt method(random)}.

{phang}{opt antithetics} specifies that antithetic draws are to be used.  The antithetic
draw for the J-1 vector uniform random variables, {it:x}, is 1-{it:x}.

{phang}{opt initbhhh(#)} requests the BHHH algorithm for the initial 
{it:#} optimization steps.  This option is the only way to use the BHHH 
algorithm along with other optimization techniques.  The algorithm 
switching feature of {cmd:ml}'s {opt technique()} option cannot include 
{opt bhhh}.

{marker maximize_options}{...}
{dlgtab:Max options}
{phang}
{it:maximize_options}:
{opt dif:ficult},
{opt tech:nique(algorithm_spec)},
{opt iter:ate(#)},
[{cmdab:no:}]{opt lo:g},
{opt tr:ace},
{opt grad:ient},
{opt showstep},
{opt hess:ian},
{opt shownrtolerance},
{opt tol:erance(#)},
{opt ltol:erance(#)},
{opt gtol:erance(#)},
{opt nrtol:erance(#)},
{opt nonrtol:erance},
{opt from(init_specs)}; see {help maximize}.  

{pmore}
The following options may be particularly useful in obtaining 
convergence with {cmd:asmprobit}: {opt difficult}, 
{opt technique(algorithm_spec)}, {opt nrtolerance(#)}, 
{opt nonrtolerance}, {opt from(init_specs)}.

{pmore}
If {opt technique()} contains more than one algorithm specification, 
{opt bhhh} cannot be one of them.  In order to use the BHHH algorithm in 
conjunction with another algorithm, use the {opt initbhhh()} option and 
specify the other option in {opt technique()}.


{title:Examples}

{phang}{cmd:. webuse travel}{p_end}

{phang}{cmd:. asmprobit choice traveltime termtime, casevars(income) case(id)}
	{cmd:alternatives(mode)}
{p_end}


{title:Also see}

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

{psee}
Online:  {help asmprobit postestimation};{break}
{help estcom}, {help postest}; {helpb clogit},
{helpb constraint}, {helpb glogit}, {helpb logistic}, {helpb logit}, 
{helpb mlogit}, {helpb mprobit}, {helpb nlogit},
{helpb ologit}, {helpb slogit}, {helpb svy}, {helpb xi}
{p_end}