Estimate a Component GARCH Model

[Yu Hu]

I'm trying to estimate a component GARCH model using Stata, but it's not a built-in function. I think it can be done by either modifying -arch- command or writing my own codes and using -ml-. I've tried to use -ml- to estimate an ARCH model (because its log likelihood function is simpler), and my codes are below:

sysuse sp500
gen r = change/close[_n-1]
tsset date

program arch

args todo b lnfj
tempvar xb
quietly generate double `xb'= `b'[1,1]
quietly replace `lnfj' = -0.5*ln(2*_pi)-0.5*ln(`b'[2,1]+`b'[2,2]* ///
(L.r- L.`xb')^2)-0.5*(r-`xb')^2/(`b'[2,1]+`b'[2,2]*(L.r-L.`xb')^2)
end

ml model gf0 arch (r=) /sigma
ml check

Then Stata reports error "could not find feasible values r(491)". Any suggestion?

[Yu Hu]

Solved.

[Laura Endres]

Solved.

Hey Yu,
would you mind posting your solution code? I've ran into a similar problem.
All the best
Laura