Page 5 of 9

Re: DCCGARCH11

Posted: Thu May 14, 2015 6:10 pm
by shuangqing
Hi trubador. I have two series of interest rate and exchange rate, and would like to use dccgarch11 add-in to know the volatility spillovers between interest rate and exchange rate, and the dynamic correlation coefficients of them. Can you show me how to put in the four empty boxes, they are “return series”, “exogenous variables in the mean equation”, “number of AR lags to be used in mean equations”, and “exogenous variables in the variance equation”. Thanks much.

Re: DCCGARCH11

Posted: Wed May 20, 2015 12:19 am
by George1993
How to use this add-in to calculate the covariance matrix of errors?

Re: DCCGARCH11

Posted: Thu May 21, 2015 7:41 am
by shuangqing
where is trubador? :cry:

Re: DCCGARCH11

Posted: Sat May 23, 2015 2:52 pm
by trubador
George1993 wrote:How to use this add-in to calculate the covariance matrix of errors?

You can generate the covariance series by exploiting their statistical relationship to variance and correlation series.

Re: DCCGARCH11

Posted: Sat May 23, 2015 2:56 pm
by trubador
shuangqing wrote:Hi trubador. I have two series of interest rate and exchange rate, and would like to use dccgarch11 add-in to know the volatility spillovers between interest rate and exchange rate, and the dynamic correlation coefficients of them. Can you show me how to put in the four empty boxes, they are “return series”, “exogenous variables in the mean equation”, “number of AR lags to be used in mean equations”, and “exogenous variables in the variance equation”. Thanks much.

Have you seen the documentation that comes with the add-in?

Re: DCCGARCH11

Posted: Mon May 25, 2015 1:33 am
by alexlooyc
Hi Trubador, I am trying to find out the DCC for the tourism demand for singapore, south korea and thailand from Indonesia. However, my estimates from the DCC-GARCH(1,1) are not making sense. I get a single result instead of a time-varying correlation coefficient.Am i estimating it correctly using the add in?

For the mean eqn; I added 11 dummy monthly variables and AR(2) as encouraged by past literature.

I have attached the workfile too!Pls assist!Thanks!

Re: DCCGARCH11

Posted: Mon May 25, 2015 2:03 am
by trubador
alexlooyc wrote:Hi Trubador, I am trying to find out the DCC for the tourism demand for singapore, south korea and thailand from Indonesia. However, my estimates from the DCC-GARCH(1,1) are not making sense. I get a single result instead of a time-varying correlation coefficient.Am i estimating it correctly using the add in?

For the mean eqn; I added 11 dummy monthly variables and AR(2) as encouraged by past literature.

I have attached the workfile too!Pls assist!Thanks!


The output (i.e. dccout01) clearly indicates that the optimization was not successful. If you carefully examine your univariate GARCH estimates, you should see that the GARCH effects are very weak.

Please read the previous posts in this thread. We have had similar discussions before.

Re: DCCGARCH11

Posted: Thu May 28, 2015 5:34 am
by shuangqing
trubador wrote:
shuangqing wrote:Hi trubador. I have two series of interest rate and exchange rate, and would like to use dccgarch11 add-in to know the volatility spillovers between interest rate and exchange rate, and the dynamic correlation coefficients of them. Can you show me how to put in the four empty boxes, they are “return series”, “exogenous variables in the mean equation”, “number of AR lags to be used in mean equations”, and “exogenous variables in the variance equation”. Thanks much.

Have you seen the documentation that comes with the add-in?

Now I have got the documentation,and settled the problem. Think you, trubador.

Re: DCCGARCH11

Posted: Thu May 28, 2015 9:19 am
by Ivan10
Hi Trubador,

I have one question. I am using two time series in the add-in.

One of the two is normally distributed, the other is not (given by Jarque-Bera)

Do you suggest using Normal or T_Stud distribution for this specific case? Thank you

Re: DCCGARCH11

Posted: Thu May 28, 2015 1:40 pm
by trubador
Ivan10 wrote:Hi Trubador,

I have one question. I am using two time series in the add-in.

One of the two is normally distributed, the other is not (given by Jarque-Bera)

Do you suggest using Normal or T_Stud distribution for this specific case? Thank you

There is nothing specific here. These distributions are fit to residuals, not to the series. So, you can estimate the model with the assumption of normal distribution and then check the (standardized) residuals to verify this. You can also test alternative empirical distributions to identify the best fit.

Re: DCCGARCH11

Posted: Sat Jun 06, 2015 5:20 am
by markbaseball
Dear trubador:
I'm trying to use dcc add-in to perform my reaserch ,but I theta1 theta2 can't optimize, but I don't know why ?
I have check stationary and time trend.

Re: DCCGARCH11

Posted: Tue Jun 09, 2015 6:17 am
by trubador
markbaseball wrote:Dear trubador:
I'm trying to use dcc add-in to perform my reaserch ,but I theta1 theta2 can't optimize, but I don't know why ?
I have check stationary and time trend.

Have you read the previous posts? For instance:
viewtopic.php?f=23&t=9677&start=15#p36929
viewtopic.php?f=23&t=9677&start=45#p42744
viewtopic.php?f=23&t=9677&start=45#p42822

Re: DCCGARCH11

Posted: Tue Jun 09, 2015 6:00 pm
by markbaseball
I have read previous post .
if I don't put any exogeonus in the model, then I can get the right theta1 and theta2.

but if I put exogenous variable,I still can't get the right theta1 and theata 2

Re: DCCGARCH11

Posted: Wed Jun 10, 2015 12:58 am
by trubador
trubador wrote:1) Since this is a two-step model, you should check the first step (i.e. estimation of univariate GARCH models) to see if everything is OK. Try alternative GARCH models.
2) Model may become ill-defined or inconsistent after including an exogenous variable or an AR term. Try dropping them or find better RHS variables.
3) Starting values of coefficients may be too far from an optimal solution. Try different initial parameter values (i.e. theta vector).
4) Correlation targeting may be too restrictive. Try unchecking this option.
5) Sample period may not be appropriate to carry out such an analysis. Try adjusting the sample.
6) Optimization algorithm may perform poorly. Try other alternatives.
7) Algorithms may fall out the domain of estimation parameters. Try optimizing the squared coefficients.

Such models are nonlinear in nature and therefore there is no guarantee that they will always converge and yield proper estimation results. It really needs "your" time and effort to get it work.

Re: DCCGARCH11

Posted: Fri Jun 12, 2015 8:32 am
by selmrog
Hi Trubador,
I used the add-in for 5 time series. I think the results are fine?! I attached the work file. My question is, if its possible to get the variance covariance matrix for the time varying correlations to use the matrix for portfolio optimization. I did get it right that the rho_12_01 is the correlation between series 1 and 2 and so on?

Thank you very much for you help!!