Page 6 of 9

### Re: DCCGARCH11

Posted: Fri Jun 12, 2015 1:14 pm
selmrog wrote: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!!

Yes, the results "appear to be" fine. However, I cannot verify or validate your model, as it requires a more thorough analysis of diagnostics.
Yes, it is possible to get covariance series (though not as a matrix). Since you already have correlation series, just save the variance series from garch estimations and then use the statistical formula of correlation.
Yes, your reading of the correlation series is correct.

### Re: DCCGARCH11

Posted: Sun Jun 14, 2015 6:33 am
Thanks very much for your feedback trubador! Your add in is very helpful btw ### Re: DCCGARCH11

Posted: Mon Jun 15, 2015 3:36 am
Hey Trubador,
another question regarding the covariance matrix. First, my thesis deals with portfolio optimisation and I want to include the time varying correlation into my optimisation model. I figured out, with your help, how to get the covariance series, however I need the matrix to include it into my optimisation model. Do you have any idea how to generate such a time varying covariance matrix with the output of the add in or is it not possible at all?

### Re: DCCGARCH11

Posted: Mon Jun 15, 2015 6:27 am
Unfortunately, it would require a multidimensional array (mxnxt), which EViews does not support at the moment. However, if your optimization model takes a single (nxn) covariance matrix, then you can create one using the final values from the estimation.

### Re: DCCGARCH11

Posted: Mon Jun 15, 2015 7:26 am
Dear Trubador,
thanks for your help. If I use the single nxn covariance matrix I can use the final estimations, since this model calculates the current correlation between variables as a function of past realisations of volatility within the variables as well as the correlation between the variables. Thus to capture the time-varying nature of variances and covariance's the end of the period values can be used into the portfolio optimisation model?

### Re: DCCGARCH11

Posted: Mon Jun 15, 2015 12:00 pm
Portfolio optimization is a little bit off-topic, I am afraid. All I can say is that, this is not the correct way of incorporating the time varying relationship of assets in order to obtain a dynamic portfolio. Since the covariance matrix itself is changing over time, you'll obtain a different portfolio at any given point. Final values of the covariance matrix would show you the most recent version of your portfolio, that's all.

### Re: DCCGARCH11

Posted: Wed Jul 08, 2015 2:27 pm
What is the most appropriate input to DCC GARCH- DLOG(series), Resid from ARMA modelling of dlog(series) or Resid from ARIMA modelling of price series?

While estimating DCC GARCH whether we should start with the dlog(series) and then estimate GARCH and DCC Counterpart OR we should do ARMA modelling of dlog(series) OR ARIMA modelling of price series, obtain the residuals , check diagnostics and then proceed to estimate GARCH and DCC Counterpart.
Additionally if out of the two series one reaches white noise by just doing dlog(series) and other requires ARMA modelling, then whether the following input can be given to DCC GARCH model or not - dlog(series1) and residual of ARMA modelling(series2).
And if Not , then what is the right way of proceeding further.

### Re: DCCGARCH11

Posted: Wed Jul 08, 2015 2:30 pm
If we want to calculate time varying correlation between various asset class from 2006-2009 and if our purpose is not forecasting but analysis & inference based on historical data, then out of these two options which one is appropriate
1. Estimate the Univariate GARCH for the entire time span (2006-09) and from its residuals and conditional variance estimate Dynamic correlations.
Concern:
If we proceed like this, then since the residuals were estimated based on GARCH estimate of the entire sample (2006-09), therefore the time varying correlation, say for example in 2007 will be biased, since the residuals from which it is calculated has information till 2009. Ideally the time varying correlation in 2007 should be based only on the information available till 2007 not 2009.
2. Rolling window DCC GARCH: We should keep the window rolling either on daily basis or weekly basis and then re estimate entire model (GARCH, Residuals and DCC Counterpart) for daily rolling window or weekly rolling window.
Concern:
Since we want to analyse and inference from the data based on all the available historical data till date, so doesn’t make much sense and will just increase the volume of work.

### Re: DCCGARCH11

Posted: Mon Jul 27, 2015 2:43 am
Hi trubador,
If I have two return series (r1 r2) and two exogenous variables (i1 i2), I want to let i1 be the exogenous variable of r1, and let i2 be the exogenous variable of r2.

However, if we put "Return series: r1 r2"; and "Exogenous variables in the variance equation: i1 i2", then both i1 and i2 will be exogenous variables for r1, and both i1 and i2 will be exogenous variables for r2.

How could we solve this problem? Thanks

### Re: DCCGARCH11

Posted: Thu Jul 30, 2015 11:54 am
eastlight wrote:Hi trubador,
If I have two return series (r1 r2) and two exogenous variables (i1 i2), I want to let i1 be the exogenous variable of r1, and let i2 be the exogenous variable of r2.
However, if we put "Return series: r1 r2"; and "Exogenous variables in the variance equation: i1 i2", then both i1 and i2 will be exogenous variables for r1, and both i1 and i2 will be exogenous variables for r2.
How could we solve this problem? Thanks

I do not think this is actually a problem. If you like, you can build univariate ARIMA models at the outset and then feed the residuals to dccgarch11 add-in. In that case, however, the estimation will become a 3-step procedure. Or, you can use both of the regressors for each dependent variable. If any of the regressors are orthogonal (unrelated) to the dependent variable(s), then the variance explained by that regressor will be close to zero.

### Re: DCCGARCH11

Posted: Fri Jul 31, 2015 2:15 am
trubador wrote:
eastlight wrote:Hi trubador,
If I have two return series (r1 r2) and two exogenous variables (i1 i2), I want to let i1 be the exogenous variable of r1, and let i2 be the exogenous variable of r2.
However, if we put "Return series: r1 r2"; and "Exogenous variables in the variance equation: i1 i2", then both i1 and i2 will be exogenous variables for r1, and both i1 and i2 will be exogenous variables for r2.
How could we solve this problem? Thanks

I do not think this is actually a problem. If you like, you can build univariate ARIMA models at the outset and then feed the residuals to dccgarch11 add-in. In that case, however, the estimation will become a 3-step procedure. Or, you can use both of the regressors for each dependent variable. If any of the regressors are orthogonal (unrelated) to the dependent variable(s), then the variance explained by that regressor will be close to zero.

Thanks for your help, I tried to use the first method you offered but I don't understand that very well. If the dependent variable is in the variance equation rather than the return equation, how could we fit ARIMA model to get the residuals for dccgarch11 add-in?

### Re: DCCGARCH11

Posted: Tue Jan 05, 2016 7:42 am
Hello trubador,

The add-in dcc-garch says that the use of add-in has been expired. I would like to ask you if are you gonna to extend add for next year 2016. I would really appreciate it. Thanks

### Re: DCCGARCH11

Posted: Mon Jan 11, 2016 1:53 am
Dear Trubador,
Thank you for sharing this add in. This add in is very useful for me. But this add in has been expired. Can you extend it again? I will really appreciate it

Thank you very much. ### Re: DCCGARCH11

Posted: Sun Apr 10, 2016 10:19 am
Dear Trubador,

I have a question regarding the Add-in.
I am estimating an MV GARCH DCC(using Student's t and correlation targeting, no AR or constants).

Firstly, in the dccout01 table, the results are not properly outlined and secondly I get coefficients for the constants (not only the thetas as it should be) which I do not know how to interpret.

Do you think this is a software bug or there is something wrong with the modelling.

I have attached the file so you can have a look.

Thanks ### Re: DCCGARCH11

Posted: Mon Apr 11, 2016 5:11 am
marto91 wrote:Dear Trubador, I have a question regarding the Add-in. I am estimating an MV GARCH DCC(using Student's t and correlation targeting, no AR or constants).Firstly, in the dccout01 table, the results are not properly outlined and secondly I get coefficients for the constants (not only the thetas as it should be) which I do not know how to interpret. Do you think this is a software bug or there is something wrong with the modelling.I have attached the file so you can have a look.

Yes, you are right. It seems the add-in is not fully compatible with EViews 9, which is probably due to recently introduced optimization engine. New output messages lead to a mess in the add-in's output format. I cannot promise a certain time frame, but I'll try to fix that.
The results are correct, though. You can ignore the coefficent vector c(), and take into account only the theta vectors.