st: Fwd: different mgarch vcc predicted variance results on different pc's

[Xinjia Liu <lxjliu@gmail.com>]

Hi,
I ran the same mgarch vcc command on 2 separate PC's, and got the same
identical log as follows...

. mgarch vcc (ret x_r =), arch(1) garch(1) nolog iter(100) constraint(1/7)

Varying conditional correlation MGARCH model

Sample: 9316 - 9568                                Number of obs   =       253
Distribution: Gaussian                             Wald chi2(.)    =         .
Log likelihood =  1567.101                         Prob > chi2     =         .

 ( 1)  [ARCH_ret]L.arch = -.0454344
 ( 2)  [ARCH_ret]L.garch = 0
 ( 3)  [ARCH_ret]_cons = .0001649
 ( 4)  [ARCH_x_r]L.arch = .1162334
 ( 5)  [ARCH_x_r]L.garch = .5921709
 ( 6)  [ARCH_x_r]_cons = .0000371
 ( 7)  [corr(ret,x_r)]_cons = .4823299
-------------------------------------------------------------------------------
              |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
--------------+----------------------------------------------------------------
ret           |
        _cons |    .001041   .0008262     1.26   0.208    -.0005783    .0026604
--------------+----------------------------------------------------------------
ARCH_ret      |
         arch |
          L1. |  -.0454344  (constrained)
              |
        garch |
          L1. |          0  (omitted)
              |
        _cons |   .0001649  (constrained)
--------------+----------------------------------------------------------------
x_r           |
        _cons |   .0015137   .0006759     2.24   0.025      .000189    .0028385
--------------+----------------------------------------------------------------
ARCH_x_r      |
         arch |
          L1. |   .1162334  (constrained)
              |
        garch |
          L1. |   .5921709  (constrained)
              |
        _cons |   .0000371  (constrained)
--------------+----------------------------------------------------------------
 corr(ret,x_r)|   .4823299  (constrained)
--------------+----------------------------------------------------------------
Adjustment    |
      lambda1 |   .0151112   .0158745     0.95   0.341    -.0160024    .0462247
      lambda2 |   .9475648   .0370787    25.56   0.000     .8748919    1.020238
-------------------------------------------------------------------------------

When I run the predict variance command...

. predict temp*, variance

On one PC, I got time varying variance and covariance predictions as expected...

. sum temp*

    Variable |       Obs        Mean    Std. Dev.       Min        Max
-------------+--------------------------------------------------------
temp_ret_ret |       253    .0001578    .0000147   8.88e-06   .0001649
temp_x_r_ret |       253    .0000629    .0000131    .000033   .0001827
temp_x_r_x_r |       253     .000128    .0000637   .0000953   .0007348


On the other PC, the predicted variances and covariance are wrong:
  - the variances are equal to 1
  - the covariance is time varying but not consistent with the
estimation result above

. sum temp*

    Variable |       Obs        Mean    Std. Dev.       Min        Max
-------------+--------------------------------------------------------
temp_ret_ret |       253           1           0          1          1
temp_x_r_ret |       253    .4514813    .0403912   .3386043   .5350307
temp_x_r_x_r |       253           1           0          1          1

The two computers are identical Dell Optiplex 9010, Windows7 64bit,
with the same version of Stata (MP 13.0). I was able to replicate the
results multiple times.

Does anyone have an idea of what might have caused this?

Thanks,
Xinjia
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/