[R-SIG-Finance] Problem with ugarchroll

[alexios]

Hi Owe,

I have tracked down the bug, which only affects the fGARCH model, and 
will upload to r-forge soon (should be available to download in the next 
cycle of checks/builds on r-forge). In the meantime, you can should use 
this equivalent specification:

gjr.spec = ugarchspec(variance.model = list(model = "gjrGARCH", 
garchOrder = c(2,2)),mean.model list(armaOrder=c(1,1)))

There is no reason to use the fGARCH spec for models which are
already implemented seperately (i.e. the gjr and aparch models)
since it is more highly parametrized (being an omnibus model) and
there will be a little more overhead in estimation.

Hope that helps.

Regards,
Alexios Ghalanos



On 3/13/2010 2:00 PM, Owe Jessen wrote:
> Hi,
>
> i've got a problem with ugarchroll from rgarch - can anybody help me
> explain what the following error message means and how i might
> circumvent the problem?
>
> Done!...all converged.
> Fehler in ans$z : $ operator is invalid for atomic vectors
> Zusätzlich: Warnmeldung:
> In arima(data, order = c(armap, 0, armaq), include.mean = include.mean, :
> possible convergence problem: optim gave code=1
>
> Here's some code:
> require(rgarch)
> require(quantmod)
> getSymbols("^GDAXI", src="yahoo", from=1991-01-02)
> r <- dailyReturn(GDAXI)
> r.out <- window(r, start="2007-07-01")
> n.out <- length(r.out)
> gjr.spec = ugarchspec(variance.model = list(model = "fGARCH", garchOrder
> = c(2,2),submodel="GJRGARCH"),mean.model =
> list(armaOrder=c(1,1)),distribution.model= dist)
> gjr.roll <- ugarchroll(gjr.spec,
> r,n.ahead=10,forecast.length=n.out,refit.window="moving")
>
> Thanks in advance.
>
> Owe
>