DCC model of Tse and Tsui (2002)

[sana]

Dear colleague,

I would like to estimate a “DCC” version of the Bivariate GARCH(1,1) model of Tse and Tsui (2002) and I would like to add two exogenous variables (basis1 and basis2) into the conditional variances and correlation equations.

I am using the RATS software (version 6).

The output of the program still indicates an error which is a negative sign in the GARCH effect in the correlation equation:


ρt = (1-κ1-κ2) ‾ρ + κ1 ρt-1 + κ1 ψt-1 + µβ+ t-1 + ν β- t-1

Variable Coeff Std Error T-Stat Signif
*******************************************************************************
1. C 0.796194315 0.005812909 136.97002 0.00000000
2. A 0.293297346 0.008184201 35.83702 0.00000000
3. B -0.138632717 0.011278177 -12.29212 0.00000000
4. BASISP{1} -0.477460998 0.339600804 -1.40595 0.15973963
5. BASISN{1} 1.147928315 1.173032498 0.97860 0.32777820

Please if some one has any possible comment I will be grateful.

Thanks for help.

Best regards

[luxu1983]

may you attach your code?
thank you

[TomDoan]

Dear colleague,

I would like to estimate a “DCC” version of the Bivariate GARCH(1,1) model of Tse and Tsui (2002) and I would like to add two exogenous variables (basis1 and basis2) into the conditional variances and correlation equations.

I am using the RATS software (version 6).

The output of the program still indicates an error which is a negative sign in the GARCH effect in the correlation equation:


ρt = (1-κ1-κ2) ‾ρ + κ1 ρt-1 + κ1 ψt-1 + µβ+ t-1 + ν β- t-1

Variable Coeff Std Error T-Stat Signif
*******************************************************************************
1. C 0.796194315 0.005812909 136.97002 0.00000000
2. A 0.293297346 0.008184201 35.83702 0.00000000
3. B -0.138632717 0.011278177 -12.29212 0.00000000
4. BASISP{1} -0.477460998 0.339600804 -1.40595 0.15973963
5. BASISN{1} 1.147928315 1.173032498 0.97860 0.32777820

Please if some one has any possible comment I will be grateful.

Thanks for help.

Best regards

That looks like univariate output. What happens to the uni GARCH models when you leave out the shifts in the variance? This isn't showing particularly strong "GARCH" effects (both A and B are fairly small).