I would think that the thing to do
would be 'locf' last observation
carried forward (on prices). If there
is no activity, then presume that the
price hasn't changed. You then want to
get returns. So you'll end up with a
number of zeros.
There are other issues that make intraday
garch modeling "interesting". There is
typical seasonality of the volatility
through the day. Doing it "right" is hard.
Doing it well enough to add value, might
not be terribly hard.
Pat
On 26/01/2012 18:53, Ted Byers wrote:
Thanks Alexios
It is easy enough to extract a vector of unnamed numeric values.
But that leaves the question, what is the best option for filling in missing
values with some suitable default value (which obviously will have to change
through the day).
Is there something in a related package that will provide a suitable
interpolation?
Thanks again,
Ted
-----Original Message-----
From: alexios [mailto:[email protected]]
Sent: January-26-12 1:48 PM
To: Ted Byers
Cc: [email protected]
Subject: Re: [R-SIG-Finance] rugarch and missing data
Hi Ted,
1. rugarch does not handle missing values.
2. you should pass tick/intraday data as unnamed numeric (rugarch does not
specifically handle intraday formatted data, so unnamed numeric is the the
best option).
3. rmgarch is available from r-forge under the rgarch project.
and can be installed by stating that in the 'repos' argument of the
install.packages i.e.
install.packages("rmgarch",repos="http://r-forge.r-project.org";)
You'll need to manually install some of the required dependencies.
Regards,
Alexios
On 26/01/2012 18:32, Ted Byers wrote:
Hi all,
I seek enlightenment.
I am beginning to study rugarch, to get a better sense of what it can
do.
The data I am tying first is tick data for futures contracts, but I
suppose any tick data will have similar issues.
Here is how I am setting up my data:
x = read.table("quotes_M11.dat", header = FALSE, sep="\t", skip=0)
dt<-sprintf("%s %04d",x$V2,x$V4)
dt<-as.POSIXlt(dt,format="%Y-%m-%d %H%M")
y<- data.frame(dt,x$V5)
colnames(y)<- c("tickdate","price")
z<- xts(y[,2],y[,1])
alpha<- to.minutes(z, OHLC=TRUE)
colnames(alpha)<- c("Open","High","Low","Close")
So, x has the raw tick data. NB: I can get the same data, with the
same structure, using SQL from my DB, but didn't want to complicate
things. This data file has but a small subsample of all the data I can
work
with.
dt gets the date and time data into a single field, as a datetime
object that to.minutes will accept.
y gets a data.frame in which the first column is the datetime and the
second is the price, and z gets an xts object from that.
Finally, alpha gets one minute OHLC data. Not surprisingly, alpha
does not have values for every minute of every day for which there is
data in quotes_M11.dat
I use alpha successfully for an analysis I do using rollapply. That
works great, and it doesn't seem to care that there are minutes without
values.
(I don't know that package well enough to know why, so a little
enlightenment would be great).
The problem I have is this.
spec = ugarchspec(variance.model = list(model = "eGARCH", garchOrder =
c(1, 1)), mean.model = list(armaOrder = c(1, 1), arfima = FALSE),
distribution.model = "std")
fit = ugarchfit(spec = spec, data = alpha$Close, out.sample = 0,
solver = "solnp", solver.control = list(trace = 0))
Error in if (all(dte[, i]<= 12)) m = i :
missing value where TRUE/FALSE needed
In addition: There were 50 or more warnings (use warnings() to see the
first
50)
My reaction to this is, well of course there's missing values. The
warnings are about NAs being produced (In FUN(newX[, i], ...) : NAs
introduced by coercion).
So, I understand the error message and why it happens. The question
is, "What can I do about it?"
Are any of the steps I show in my data manipulation unnecessary? (I
expect that they can be condensed into fewer statements, but I like to
see precisely what each step does before I condense the code.)
Is there another step in my data manipulation that I need to add?
Can any of the methods in the rugarch handle raw tick data directly?
As an aside, I saw mention of rmgarch in the documentation, but when I
select "Install packages" from the Packages submenu, I do not see it
listed.
Is it not yet available? If so, is rgarch the principle alternative
for multivariate analyses (and if so, with what caveats)?
Any enlightenment beyond what I found in the manual would be greatly
appreciated.
Thanks
Ted
[[alternative HTML version deleted]]
_______________________________________________
[email protected] 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.
_______________________________________________
[email protected] 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.
--
Patrick Burns
[email protected]
http://www.burns-stat.com
http://www.portfolioprobe.com/blog
twitter: @portfolioprobe
_______________________________________________
[email protected] 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.
