[R-SIG-Finance] ugarchspec: external regressors

[FAKIR CHARLES]

Hi Alexios,

Thanks for all your hardwork, I think many would agree with if I say you
are doing a great job.

As for me, I have less than 6 months experience with R, so I already feel
dumb, and when the solution is so simple, well you can imagine how I feel.
Anyway I got my ugarchspec with the external regressor, and now will have a
few days trying to figure out, my ugarchfit problem:

> fit1X=ugarchfit(data=close.djdr,spec=spec1X)
Error in temp$h : $ operator is invalid for atomic vectors
In addition: Warning message:
In .makefitmodel(garchmodel = "sGARCH", f = .sgarchLLH, T = T, m = m,  :
rugarch-->warning: failed to invert hessian

So, don't be surprised if this thread continues.

Regards
Fakir Charles

On Tue, May 24, 2016 at 1:36 AM, alexios galanos <alexios at 4dscape.com>
wrote:

> I think the error message is quite informative:
> "...unused argument (external.regressors...)
>
> The "external.regressors" goes either in the mean.model or
> variance.model list.
> You are passing it outside, so you get an error. I did mention to you
> when you
> emailed me offline that it might be useful to carefully check the
> ugarchspec specification and what you are passing.
>
> A quick "arguments" check usually helps:
>
> >args(ugarchspec)
> function (
> variance.model = list(model = "sGARCH", garchOrder = c(1, 1),
> submodel = NULL, external.regressors = NULL, variance.targeting = FALSE),
> mean.model = list(armaOrder = c(1, 1), include.mean = TRUE,
> archm = FALSE, archpow = 1, arfima = FALSE, external.regressors = NULL,
> archex = FALSE),
> distribution.model = "norm", start.pars = list(), fixed.pars = list(), ...)
>
>
> Alexios
>
> On 23/05/2016 16:02, FAKIR CHARLES wrote:
> > Hi All,
> >
> > I am running rugarch Version: 1.3-5 and am trying to simulate the paper:
> > “High-low range in GARCH models of stock return volatility” SSRN:
> > http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2753336, for my
> thesis.
> >
> > I am able to run sGARCH with no problems. The paper though, calls to
> > “estimate a GARCH(0,1) model with an exogenous variable” This is where it
> > goes all pear shape. I have modelled the volatility using the TTR
> package:
> > “train.vGK <- volatility(train.ohlc, calc="garman")
> > train.vParkinson <- volatility(train.ohlc, calc="parkinson")”
> >
> > The result is an xts object, as expected considering my input class is
> xts,
> > and the dimensions after some adjustments:
> >
> >> str(close.djdr)
> > An ‘xts’ object on 1993-01-15/2009-04-24 containing:
> >   Data: num [1:4100, 1] 0.000979 0.001161 -0.005788 -0.004309 0.003387
> ...
> >  - attr(*, "dimnames")=List of 2
> >   ..$ : NULL
> >   ..$ : chr "daily.returns"
> >   Indexed by objects of class: [Date] TZ: UTC
> >   xts Attributes:
> >  NULL
> >
> >> str(train.vParkinson)
> > An ‘xts’ object on 1993-01-15/2009-04-24 containing:
> >   Data: num [1:4100, 1] 0.0866 0.0853 0.0875 0.088 0.0772 ...
> >  - attr(*, "dimnames")=List of 2
> >   ..$ : NULL
> >   ..$ : chr "vParkinson"
> >   Indexed by objects of class: [Date] TZ: UTC
> >   xts Attributes:
> >  NULL
> >
> > When I try to run the ugarchspec, it fails miserably:
> >> spec1X=ugarchspec(variance.model=list(model="sGARCH"),
> > + external.regressors = train.vParkinson[,1],  #What is the correct form
> > here? I have lagged it: train.xreg = lag(train.vParkinson,1)
> > + mean.model=list(armaOrder=c(0,0)))
> > Error in .local(variance.model, mean.model, distribution.model,
> > start.pars,  :
> >   unused argument (external.regressors = c(0.0866063373770964,
> > 0.0853297963049413, 0.0875414180301314, 0.0880414499227675,
> > 0.0771631965517572, 0.0707091044790502, 0.0778057119517414,
> > 0.0775765066940466, 0.077820690394834, 0.0752388231276861,
> > 0.072436404409338, 0.07452357235863, 0.0710133687112722,
> > 0.0817565181849798, 0.0900172823007909, 0.0930080887667925,
> > 0.0896830143657568, 0.089774572544706, 0.0904325308768051,
> > 0.092930283058663, 0.0964456153316399, 0.123570155397932,
> > 0.125244790062196, 0.138922401992731,
> > 0.135292796408894, 0.134903364202769, 0.136922606737311,
> 0.138012028037638,
> > 0.136775934192342, 0.136172719004911, 0.136364919382056,
> 0.122106445616322,
> > 0.121111651605594, 0.0962623729322267, 0.0981395485851856,
> > 0.117961280078613, 0.114364914549088, 0.112811619226829, 0.1139844134944,
> > 0.120975655704514, 0.117871553416396, 0.106136285136668,
> 0.106057886734732,
> > 0.113735726997588, 0.108700010793231, 0.085947506860437,
> > 0.0853823006923295, 0.0834373451373239, 0.0829486747947965,
> >
> > So, I lag my exon variable
> > train.xreg = lag(train.vParkinson,1)
> > train.test = as.matrix(train.xreg, ncol = 1)
> >
> > Then set a matrix, but I don’t think this is required as I am using xts
> > class (see:
> https://stat.ethz.ch/pipermail/r-sig-finance/2014q3/012708.html)
> >
> > I have tried to change the class of train.vParkinson to matrix,
> data.frame,
> > with the square brackets, without the square brackets, read several
> threads
> > on a similar problem. What am I doing wrong?
> >
> > The data is extracted with quant mod:
> > #     1.1.1 Set start and end date for collection in YYYYMMDD (numeric)
> > format
> > train.date.start<-19930103 #Monday
> > train.date.end<-20090424 #Friday - changed the original date which was
> > 20090427
> >
> >
> > #     1.1.2 Collect historical data for DJIA Index
> > train.djd <- getYahooData("^DJI", start=train.date.start,
> > end=train.date.end, freq="daily")
> > chartSeries(train.djd[,1:5])
> >
> > close.djdr = dailyReturn(djd[,4], type="log") #daily returns close on
> close
> > close.djdr = close.djdr[-(1:9)]
> >
> > Thanks in advance.
> > Kirro
> >
> >       [[alternative HTML version deleted]]
> >
> > _______________________________________________
> > R-SIG-Finance at r-project.org mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-sig-finance
> > -- Subscriber-posting only. If you want to post, subscribe first.
> > -- Also note that this is not the r-help list where general R questions
> should go.
>
>
>

	[[alternative HTML version deleted]]