Page 2 of 2
Re: VAR E-GARCH with dummies
Posted: Mon Aug 09, 2010 7:53 am
by TomDoan
The problem with the EGARCH seems like it might be specific to the data set. Please attach it.
You're correctly interpreting the other result.
Re: VAR E-GARCH with dummies
Posted: Mon Aug 09, 2010 9:52 am
by Eric
.
Re: VAR E-GARCH with dummies
Posted: Tue Aug 10, 2010 8:22 am
by TomDoan
I converted the Tse test to a procedure, which is now posted at:
http://www.estima.com/forum/viewtopic.php?f=7&t=796
The Tse test takes the same form with other types of mean models. However, you want to use the procedure, since adapting the original program requires adjusting the value of "nslot" to allow for the number of parameters in the mean model.
It often helps with GARCH models (particularly MVGARCH) to scale up the returns from fractions to percentages, which I've done below. This seems to work properly.
Code: Select all
*
* Multivariate GARCH with two-step DCC estimator
*
cal(ppd=119,i)
open data barclaysdata.txt
data(org=obs) 1 8377 BarclaysUK BarclaysUS
set BarclaysUK = 100.0*BarclaysUK
set BarclaysUS = 100.0*BarclaysUS
*
garch(p=1,q=1,nomean,mv=cc,exp,asymmetric,derives=ccderives,rvector=uv,hmat=h,iters=400) / BarclaysUK BarclaysUS
@tsecctest(rvector=uv,hmat=h,derives=ccderives)
garch(p=1,q=1,nomean,mv=cc,asymmetric,derives=ccderives,rvector=uv,hmat=h,iters=400) / BarclaysUK BarclaysUS
@tsecctest(rvector=uv,hmat=h,derives=ccderives)
garch(p=1,q=1,nomean,mv=cc,derives=ccderives,rvector=uv,hmat=h,iters=400) / BarclaysUK BarclaysUS
@tsecctest(rvector=uv,hmat=h,derives=ccderives)
Re: VAR E-GARCH with dummies
Posted: Tue Aug 10, 2010 10:58 am
by Eric
Dear Tom,
The Tse test procedure gives this message:
The Error Occurred At Location 0297 of TSECCTEST
Line 32 of TSECCTEST
## MAT14. Non-invertible Matrix. Using Generalized Inverse for SYMMETRIC.
## MAT17. Can't Use Row Range of 1 to 0 in Left out of table operation
The Error Occurred At Location 0394 of TSECCTEST
Line 39 of TSECCTEST
Thanks,
Eric
Re: VAR E-GARCH with dummies
Posted: Tue Aug 10, 2010 12:23 pm
by TomDoan
What version of RATS are you using? Warwick should have version 7.3.
You don't have to do anything about "nslot". That was the old program. The procedure figures out where everything is based upon your inputs. (The CC parameters are always at the end).