GARCH estimation with some fixed parameters

[arash]

Hi there,
I want to estimate a general GARCH(1,1) model, with some fixed parameters in its variance formula. Meanwhile, I have faced a problem which made me to send this post here and ask for your kind advise. Actually I don't know how to define my variance formula and also how to fit one parameter to be constant through the GARCH parameters estimation of my model. For example in a model like below:
:equation: GARCH = C(2) + C(3)*RESID(-1)^2 + C(4)*GARCH(-1)
I want to fix one of the parameters C(2) or C(3) or C(4), then estimate other not fixed parameters. In an example to clarify the stuff, I want to estimate the below model:
:equation: GARCH = 0.75 + C(3)*RESID(-1)^2 + C(4)*GARCH(-1)
And then I want to estimate C(3) and C(4).
Many thanks in advance. I appreciate your kind cooperation and looking forward to hearing from you soon.
Best Regards,
Arash

[EViews Gareth]

I don't believe there is a way to do this, although it may be possible to specify a LogL object with the likelihood for the particular likelihood for your GARCH model.

[Gene]

Hi there,
I want to estimate a general GARCH(1,1) model, with some fixed parameters in its variance formula. Meanwhile, I have faced a problem which made me to send this post here and ask for your kind advise. Actually I don't know how to define my variance formula and also how to fit one parameter to be constant through the GARCH parameters estimation of my model. For example in a model like below:
:equation: GARCH = C(2) + C(3)*RESID(-1)^2 + C(4)*GARCH(-1)
I want to fix one of the parameters C(2) or C(3) or C(4), then estimate other not fixed parameters. In an example to clarify the stuff, I want to estimate the below model:
:equation: GARCH = 0.75 + C(3)*RESID(-1)^2 + C(4)*GARCH(-1)
And then I want to estimate C(3) and C(4).
Many thanks in advance. I appreciate your kind cooperation and looking forward to hearing from you soon.
Best Regards,
Arash

What you want to do is not easy, but can be done in Logl object. If you want to attempt to do this you should use one of the example programs as a starting point. The directory Eviews6\Example Files\sample Programs\LogL contains several garch programs, each program sets up and estimates a different garch model using the Maximum likelihood object, Logl. You can modify one of the programs to fit your needs.

For example arch_1t.prg estimates a GARCH(1,1) model with t-distributed errors. The line

Code: Select all

ll1.append sig2 = omega(1)+alpha(1)*res(-1)^2 +beta(1)*sig2(-1)

specifies the conditional variance equation. If you change this line to

Code: Select all

ll1.append sig2 = .025+alpha(1)*res(-1)^2 +beta(1)*sig2(-1)

the resulting program estimates a GARCH(1,1) model with t-distribution error and coefficient omega(1) restricted to .025.

[arash]

Dear Gene,
Thank you in advance for your helpful cooperation. But this example does not work in eviews and I don't know why! But I think it has some problems in this part below, especially the first line:

' set up GARCH likelihood
logl ll1
ll1.append @logl logl
ll1.append res = y-mu(1)
ll1.append sig2 = omega(1)+alpha(1)*res(-1)^2 +beta(1)*sig2(-1)
ll1.append z = res^2/sig2/(tdf(1)-2) + 1
ll1.append logl = @gammalog((tdf(1) + 1)/2) - @gammalog(tdf(1)/2) - log(!pi)/2 - log(tdf(1) - 2)/2 - log(sig2)/2 - (tdf(1)+1)*log(z)/2


When I try to run this sample with my data, it will show me an error contain this message: "Not available in Eviews Basics."
Could you or anyone else please help me?
Regards,
Arash

[EViews Gareth]

Which version of EViews are you using?

[arash]

Which version of EViews are you using?

I use eviews 5.1 .
I found an other problem in this example:
' get starting values from Gaussian ARCH
equation eq1
eq1.arch y c
show eq1.output

if you run this part of the example program, you will receive an error. The corrected formula is as below:

equation eq1.arch(1,1) y c
show eq1.output

But I couldn't correct the part that "set up GARCH likelihood" yet.