Simultaneous equation 3SLS post estimation test

[fai ot]

Dear All,

I am running a 3SLS regression and I used for post estimation test the " lmhreg3 " command to test Heteroscedasticity. I don't know exactly how to correct Heteroscedasticity since Robust option is not allowed with reg3 command.

Many thanks for you help.
Fairouz

=================================================
* System Heteroscedasticity Tests (3sls)
=================================================
*** Single Equation Heteroscedasticity Tests:
Ho: Homoscedasticity - Ha: Heteroscedasticity

Eq. WomenEmp: Engle LM ARCH Test: E2 = E2_1= 75.5039 P-Value > Chi2(1) 0.0000
Eq. WomenEmp: Hall-Pagan LM Test: E2 = Yh =222.3831 P-Value > Chi2(1) 0.0000
Eq. WomenEmp: Hall-Pagan LM Test: E2 = Yh2 =376.9723 P-Value > Chi2(1) 0.0000
Eq. WomenEmp: Hall-Pagan LM Test: E2 = LYh2= 88.2488 P-Value > Chi2(1) 0.0000
------------------------------------------------------------------------------
Eq. ODA_G: Engle LM ARCH Test: E2 = E2_1= 48.8804P-Value > Chi2(1)0.0000
Eq. ODA_G: Hall-Pagan LM Test: E2 = Yh= 18.2014P-Value > Chi2(1)0.0000
Eq. ODA_G: Hall-Pagan LM Test: E2 = Yh2= 17.8145P-Value > Chi2(1)0.0000
Eq. ODA_G: Hall-Pagan LM Test: E2 = LYh2= 17.6258P-Value > Chi2(1)0.0000
------------------------------------------------------------------------------
*** Overall System Heteroscedasticity Tests:
Ho: No Overall System Heteroscedasticity

- Breusch-Pagan LM Test = 444.3880 P-Value > Chi2(1) 0.0000
- Likelihood Ratio LR Test =1036.5641 P-Value > Chi2(1) 0.0000
- Wald Test = 1.26e+06 P-Value > Chi2(1) 0.0000
------------------------------------------------------------------------------

[Andrew Musau]

reg3 allows -vce(robust)-

webuse klein, clear
reg3 (consump wagepriv wagegovt) (wagepriv consump govt capital1), vce(robust)

Res.:

.         . reg3 (consump wagepriv wagegovt) (wagepriv consump govt capital1), vce(robust)

Three-stage least-squares regression
------------------------------------------------------------------------------
Equation             Obs   Params         RMSE  "R-squared"      chi2   P>chi2
------------------------------------------------------------------------------
consump               22        2     1.776297      0.9388     207.70   0.0000
wagepriv              22        3     2.372443      0.8542     658.77   0.0000
------------------------------------------------------------------------------

------------------------------------------------------------------------------
             |               Robust
             | Coefficient  std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
consump      |
    wagepriv |   .8012754   .0847552     9.45   0.000     .6351582    .9673926
    wagegovt |   1.029531   .2373313     4.34   0.000     .5643704    1.494692
       _cons |    19.3559   2.752245     7.03   0.000      13.9616     24.7502
-------------+----------------------------------------------------------------
wagepriv     |
     consump |   .4026076   .2042363     1.97   0.049     .0023117    .8029034
        govt |   1.177792   .4060833     2.90   0.004     .3818834    1.973701
    capital1 |  -.0281145   .0554703    -0.51   0.612    -.1368343    .0806054
       _cons |   14.63026   12.55785     1.17   0.244    -9.982685     39.2432
------------------------------------------------------------------------------
Endogenous: consump wagepriv 
Exogenous:  wagegovt govt capital1

[fai ot]

thanks for sharing this. Unfortunately I am getting the error :
option vce() not allowed
r(198);

It might be related to the stata version. I am using stata 17

[Andrew Musau]

Should work in the current version (StataNow 18.5).

[sabrina queen]

Hi Sir Andrew Musau

I would like to get insight from you

Now, I am building a simultaneous equation, and it will be run with the reg3 syntax

along with the simultaneous equation :
Y1 = X1 + X2 + X3 + X4
Y2 = Y1 + X1 + + X4 + X5 + X6

the variables Y1 and X1 are bound to the instrument variables X7 + X8 so the syntax I created is like this

reg3 (eq1: Y1 = X1 + X2 + X3 + X4) (eq2: Y2 = Y1 + X1 + + X4 + X5 + X6), endog(Y1 X1) exog (X7 X8) first

I want to ask
1. Can the independent variable in EQ 1 also be used as an independent variable in EQ 2? such as X1 and X4. For the notes, X1 is an endogenous variable that is influenced by the instrument variable, while X4 is not.
2. I see that the instrument variables X7 and X8 affect not only X1 but also Y1, so I also make Y1 an endogenous variable that is also bound to the instrument variable, is this possible to estimate?


thank you