The RATS Software Forum

The asymptotic standard error for a correlation (under the null that the correlation is zero) is 1/sqrt(# of observations). So the t-statistics can be computed with:

Code: Select all

cmom(corr)
# variables
compute %cmom=%cmom*sqrt(%nobs)
*
disp %cmom

Tom,
Thanks for the quick reply.
After running the code, I got an estimate.
I want to know whether the estimate as per code is the standard errors OR
the square root of the number of observation which I should use to calculate the t-stats.

Thanks

That produces the t-statistics. The diagonal elements are nonsense, of course, but all the off-diagonals are the asymptotic t-statistics.

Dear Friends,

I have tried to estimate the restricted model as done by Koutmos (1996) Table 2. It does not give same results. Is there any simple way to modify the full-model code to become the restricted model. I have tried to put A(1)(3)==A(2)(3)==A(3)(3)==0.0 immediately after " nonlin b a g d rr " but it does not work.

and How to automatically calculate the impact of symmetric Innovations on volatility using RATS as shown in Table 4. Thanks.

Hi Tom,
I tried to adjust your four variable code to a bivariate system. And just changed the below codes as seen. After I run the code, I have the below results. Is something wrong here? I am confused about the results.

Code: Select all

compute n=2
*
* Define the return series
*
dec vect[series] r(n)
compute start=1, end=750
set r(1) = 100*log(fra/fra{1})
set r(2) = 100*log(ger/ger{1})
*
* Template for mean equation. This is a VAR(1)
*
equation meaneq *
# constant r(1){1} r(2){1}


Code: Select all

    Variable                        Coeff      Std Error      T-Stat      Signif
************************************************************************************
1.  B(1)(1)                       0.049233484  0.031080586      1.58406  0.11318028
2.  B(1)(2)                      -0.029537994  0.038492409     -0.76737  0.44286042
3.  B(1)(3)                       0.003853134  0.035468663      0.10863  0.91349209
4.  B(1)(4)                      -0.011483730  0.020734891     -0.55384  0.57969104
5.  B(1)(5)                      -0.014403906  0.021419092     -0.67248  0.50127830
6.  B(2)(1)                       0.069453165  0.051687750      1.34371  0.17904332
7.  B(2)(2)                       0.168135839  0.047800982      3.51741  0.00043577
8.  B(2)(3)                       0.023231342  0.053341768      0.43552  0.66318590
9.  B(2)(4)                      -0.015210762  0.037747528     -0.40296  0.68697730
10. B(2)(5)                      -0.010654883  0.035051930     -0.30397  0.76114754
11. A(1)(1)                       0.002922546  0.008504550      0.34365  0.73111326
12. A(1)(2)                       0.146621640  0.034872157      4.20455  0.00002616
13. A(1)(3)                      -0.009506499  0.023067570     -0.41212  0.68025492
14. A(2)(1)                       0.037589459  0.016418051      2.28952  0.02204915
15. A(2)(2)                       0.033118479  0.015117959      2.19067  0.02847558
16. A(2)(3)                       0.103180374  0.038044773      2.71208  0.00668630
17. D(1)                         -1.332757517  0.398151163     -3.34737  0.00081584
18. D(2)                         -0.541204025  0.302098755     -1.79148  0.07321622
19. G(1)                          0.947575690  0.009841957     96.27919  0.00000000
20. G(2)                          0.952638476  0.019601947     48.59918  0.00000000
21. RR(1,1)                       0.362829658  0.028700696     12.64184  0.00000000


Thanks,
Ibrahim

Tom,
Please ignore the results at my last post. I solved the results error and got new results as seen below for a bivariate egarch model.
For this model, "B" coefficients reflect return spillover and "A" coefficients reflect volatility spillover, right!
And B(1)(2) means own lagged retuns effect and B(1)(3) means the other variables' lagged return effect, right!
I am confused about interpreting the results!
Could you please help me?
Thank you very much.

Code: Select all

MAXIMIZE - Estimation by BFGS
Convergence in    27 Iterations. Final criterion was  0.0000095 <=  0.0000100
Usable Observations                       746
Function Value                     -2340.3786

    Variable                        Coeff      Std Error      T-Stat      Signif
************************************************************************************
1.  B(1)(1)                       0.043866259  0.030047126      1.45992  0.14431336
2.  B(1)(2)                      -0.028237331  0.032342501     -0.87307  0.38262384
3.  B(1)(3)                      -0.009105750  0.019289191     -0.47206  0.63688047
4.  B(2)(1)                       0.068639510  0.052417456      1.30948  0.19037245
5.  B(2)(2)                       0.166320047  0.038417338      4.32930  0.00001496
6.  B(2)(3)                      -0.012156675  0.013639290     -0.89130  0.37276922
7.  A(1)(1)                       0.003478053  0.009199481      0.37807  0.70537816
8.  A(1)(2)                       0.143448536  0.036151163      3.96802  0.00007247
9.  A(1)(3)                      -0.006822461  0.024400134     -0.27961  0.77977863
10. A(2)(1)                       0.037727871  0.015439911      2.44353  0.01454440
11. A(2)(2)                       0.032639706  0.015393948      2.12029  0.03398120
12. A(2)(3)                       0.103701708  0.037852261      2.73964  0.00615058
13. D(1)                         -1.375944773  0.412245282     -3.33768  0.00084480
14. D(2)                         -0.538287971  0.352336195     -1.52777  0.12657014
15. G(1)                          0.947938662  0.011011507     86.08619  0.00000000
16. G(2)                          0.952677816  0.018517041     51.44871  0.00000000
17. RR(1,1)                       0.362829054  0.040470315      8.96531  0.00000000


ibrahim wrote:Tom,
Please ignore the results at my last post. I solved the results error and got new results as seen below for a bivariate egarch model.
For this model, "B" coefficients reflect return spillover and "A" coefficients reflect volatility spillover, right!
And B(1)(2) means own lagged retuns effect and B(1)(3) means the other variables' lagged return effect, right!
I am confused about interpreting the results!
Could you please help me?
Thank you very much.

The B's are the mean model coefficients. As it's written, B(x)(1) is the intercept. B(x)(2) is the coefficient on the lag of the 1st variable and B(x)(3) is on the second, so B(1)(3) is the spillover in the first equation and B(2)(2) is the spillover in the second.

The A's have a similar pattern--A(x)(1) is the intercept, ...

Thank you so much Tom,
I really appreciate your help.

One more question about bivariate VAR(1) EGARCH results. After checking diagnostic tests, LBQ for u(1) is insignificant but LB-Q for u(2) is significant. Although I specified different VAR models, diagnostis test has never changed. Actually, I don't know what to do?

Thank you very much.
Ibrahim

ibrahim wrote:Thank you so much Tom,
I really appreciate your help.

One more question about bivariate VAR(1) EGARCH results. After checking diagnostic tests, LBQ for u(1) is insignificant but LB-Q for u(2) is significant. Although I specified different VAR models, diagnostis test has never changed. Actually, I don't know what to do?

Thank you very much.
Ibrahim

I'm not sure what you mean by "I specified different VAR models". Did you increase the number of lags? That would be the only thing likely to fix the Q statistic. And how significant is it?