The RATS Software Forum

Hello, i run i run a DCC GARCH model using this code: garch(p=1,q=1,mv=dcc) / USA Canada France And i get this output.

MV-GARCH, DCC - Estimation by BFGS
Convergence in 19 Iterations. Final criterion was 0.0000000 <= 0.0000100
Daily(5) Data From 2020:03:11 To 2021:05:27
Usable Observations 317
Log Likelihood NA

Variable Coeff Std Error T-Stat Signif
************************************************************************************
1. Mean(1) 0.000985191 0.000514254 1.91577 0.05539470
2. Mean(2) 0.001600454 0.000376641 4.24928 0.00002145
3. Mean(3) -0.000057864 0.000187580 -0.30847 0.75772142
4. C(1) 0.000133110 0.000004175 31.88616 0.00000000
5. C(2) 0.000222734 0.000004540 49.06558 0.00000000
6. C(3) 0.000125230 0.000001676 74.72825 0.00000000
7. A(1) 0.045652019 0.001874274 24.35718 0.00000000
8. A(2) 0.031355016 0.001369989 22.88706 0.00000000
9. A(3) 0.093556171 0.010296647 9.08608 0.00000000
10. B(1) 0.559495261 0.014000800 39.96166 0.00000000
11. B(2) 0.478044563 0.007876932 60.68918 0.00000000
12. B(3) 0.503694765 0.016771631 30.03254 0.00000000
13. DCC(1) 0.379747476 0.009036136 42.02543 0.00000000
14. DCC(2) 1.534715948 0.037778106 40.62448 0.00000000

I need now to obtain the graphic of the conditional correlation. I find this code on the User's Guide of rats, but it doesn't work for me. Please does someone can help me.
*
set USACanada = %cvtocorr(hh(t))(1,2)
set USAFrance = %cvtocorr(hh(t))(1,3)
set CanadFrance = %cvtocorr(hh(t))(2,3)
*
spgraph(vfields=3,footer="Conditional Correlations")
graph(header="USA with Canada",min=-1.0,max=1.0)
# USACanada
graph(header="USA with France",min=-1.0,max=1.0)
# USAFrance
graph(header="Canada with France",min=-1.0,max=1.0)
# CanadaFrance
spgraph(done)

You're missing the HMATRICES=HH option on the GARCH instruction.

However, you also have a rather serious problem with your estimates: the DCC coefficients are out-of-range. Is it possible that you have a clear structural break in the data?

I rerun my model for UK and Canada and it's ok for the DCC(1) and Dcc(2) they don't exceed unit. But I am not able to obtain the graphic of the Conditional correlation even though i integrate the HMATRICES=HH option on the GARCH instruction. This is what I get
*
set UKCand = %cvtocorr(hh(t))(1,2)
## MAT15. Subscripts Too Large or Non-Positive
Error was evaluating entry 318

Try

set UKCand %regstart() %regend() = %cvtocorr(hh(t))(1,2)

It looks like your final data point has an NA.