Hansen-J Test STATA

Hansen-J test - STATA help gmm postestimation

dialogs: predict estat also see: gmm

-------------------------------------------------------------------------------

 Title

[R] gmm postestimation -- Postestimation Postestimation tools for gmm

Description

The following postestimation command is of special interest interest after after gmm:

command

description

  ------------------------------------------------------------------------estat overid

perform test of overidentifing overidentifing restrictions

  -------------------------------------------------------------------------

The following standard postestimation commands are also availa!le:

command

description

  ------------------------------------------------------------------------estat estimates lincom

"#$ cataloging estimation res%lts point estimates& standard errors& errors& testing& and inference for linear com!inations of coe'cients

nlcom

point estimates& standard errors& errors& testing& and inference for nonlinear com!inations of coe'cients

predict

resid%als

predictnl

point estimates& standard errors& errors& testing& and inference inference for generali(ed predictions

test

)ald tests of simple and composite linear hpotheses

testnl

)ald tests of nonlinear hpotheses

  -------------------------------------------------------------------------

Special-interest postestimation command

estat overid reports Hansen*s J statistic& which is %sed to determine the validit of the overidentifing restrictions in a +,, model .f the model is correctl speci/ed in the sense that $0(1i*%1i2!34 5 6& then the sample analog to that condition sho%ld hold at the estimated val%e of  ! Hansen*s J statistic is valid onl if the weight matri7 is optimal& meaning that it e8%als the inverse of the covariance matri7 of the moment conditions Therefore& estat overid onl reports Hansen*s J statistic after two-step or iterated estimation& or if o% speci/ed winitial2matname3 when calling gmm .n the latter case& it is o%r responsi!ilit to determine the validit of the J statistic

Snta7 for predict

predict [tpe] newvar [if] [in] [& e8%ation29e8noe8name3]

predict [tpe] 0st%!;newvar1< newvar1=  newvar184 [if] [in]

Resid%als are availa!le !oth in and o%t of sample> tpe predict  if  e2sample3  if wanted onl for the estimation sample

?o% specif one new varia!le and 2optionall3 e8%ation23& or o% specif st%!; or 8 new varia!les& where 8 is the n%m!er of moment  

e8%ations

,en%

Statistics @ Postestimation @ Predictions& resid%als& etc

ption for predict

 

B------B ----B ,ain B-------------------------------------------------------------

e8%ation29e8noe8name3 speci/es the e8%ation for which resid%als are desired Specifing e8%ation29<3 indicates that the calc%lation is to !e made for the /rst moment e8%ation Specifing e8%ation2demand3 wo%ld indicate that the calc%lation is to !e made for the moment e8%ation named demand& ass%ming there is an e8%ation named demand in the model

.f o% specif one new varia!le name and omit e8%ation23& res%lts are the same as if o% had speci/ed e8%ation29<3

Cor more information on %sing predict after m%ltiple-e8%ation estimation commands& see [R] predict

Snta7 for estat overid

estat overid

,en%

Statistics @ Postestimation @ Reports and statistics

$7amples

  Set%p  we!%se s%pDem  glo!al demand demand: 8%antit - 07!<:price pcompete income4  

0!64  glo!al s%ppl s%ppl: 8%antit - 07!=:price praw4 - 0c64  gmm 2Edemand3 2Es%ppl3& wmatri72ro!%st3 inst2demand:praw pcompete income3 inst2s%ppl:praw pcompete income3 winit2%nadF&indep3 deriv2
!tain resid%als for s%ppl e8%ation  predict rhat& e8%ation2s%ppl3

!tain resid%als for all e8%ations& storing them in do%!le-precision  predict do%!le r;

Same as a!ove  predict do%!le s< do%!le s=

Test whether the overidentifing restrictions are valid  estat overid

Saved res%lts

estat overid saves the following in r23:

  Scalars

r2J3

Hansen*s J statistic

r2J1df3 J statistic degrees of freedom r2J1p3

J statistic p-val%e

Also see

,an%al: [R] gmm postestimation

Help: [R] gmm