Scaling risk for irregularly spaced time series?
Also be aware that while the square of time may hold, you need to start thinking about the whole day once you go to sub daily intervals. Should you include the time from 5pm to the open of the asian markets? If you do your result will be artifically low. If not you will exclude data that could be informative. Sent via BlackBerry from T-Mobile -----Original Message----- From: "Brian G. Peterson" <brian at braverock.com> Date: Fri, 10 Oct 2008 16:28:45 To: Shane Conway<shane.conway at gmail.com> Cc: <r-sig-finance at stat.math.ethz.ch>; Chiquoine, Ben<BChiquoine at tiff.org> Subject: Re: [R-SIG-Finance] Scaling risk for irregularly spaced time series? Eric Zivot has previously posted some research to this list about intraday implied and realized volatility. Scaling volatility using the square root of time rule assumes independent observations, something which is often not true on shorter time spans. As for your question on irregularly spaced data, yes, in most cases, some method of regularizing the data is used to create a regular series to do other scaling calculations on. While this can introduce problems, as long as you're aware of things like day/week boundaries, then you're usually OK. I definitely suggest that you take a look at some of Eric's earlier posts on this topic, and maybe some of the other literature on intraday volatility. Regards, - Brian -- http://braverock.com/brian Ph: 773-459-4973 IM: bgpbraverock
Shane Conway wrote:
Thanks! That's my problem: I can't rely on the observations being evenly spaced (hence my reference to "irregularly spaced"). Do most people interpolate their data so that it's regularly spaced first, and if so, won't that also bias the calculation? Any suggestions for how to produce a daily volatility calculation using an irregularly spaced intraday time series? I'm not married to using the square-root rule if there's a better alternative... On Fri, Oct 10, 2008 at 5:06 PM, Chiquoine, Ben <BChiquoine at tiff.org> wrote:
I believe Davids suggestion is correct if your 8 hours are consecutive.
This may be pointing out the obvious but if your observations are
unevenly separated throughout the day your return series will not be
hourly and if there is mean reversion or momentum (evidence of both have
been found in fx data depending on the frequency of observation) your
results will be biased. Unfortunately I don't have a better way for you
to approach the problem this is just a heads up
Good luck,
Ben
-----Original Message-----
From: r-sig-finance-bounces at stat.math.ethz.ch
[mailto:r-sig-finance-bounces at stat.math.ethz.ch] On Behalf Of
davidr at rhotrading.com
Sent: Friday, October 10, 2008 4:52 PM
To: Shane Conway; r-sig-finance at stat.math.ethz.ch
Subject: Re: [R-SIG-Finance] Scaling risk for irregularly spaced time
series?
You basically said it: scale by the 'activity'.
If you are measuring activity for 8 hours and that is your day,
then sigma{1 hour} * sqrt(8) is your daily vol,
assuming lots of untrue things, of course ;-)
-- David
-----Original Message-----
From: r-sig-finance-bounces at stat.math.ethz.ch
[mailto:r-sig-finance-bounces at stat.math.ethz.ch] On Behalf Of Shane
Conway
Sent: Friday, October 10, 2008 3:41 PM
To: r-sig-finance at stat.math.ethz.ch
Subject: [R-SIG-Finance] Scaling risk for irregularly spaced time
series?
I'm working with intraday FX price data (primarily hourly bars). I
want to scale my volatility calculations up to the daily level.
Ordinarily I would us the square-root-of-time rule and multiple by the
sqrt(T).
The question is: how do people deal with this scaling factor when the
time series is irregularly spaced? If I apply sqrt(24) for hourly
data but I only have 8 hours of data (for instance), my calculation
will be way off.
Thanks,
Shane
_______________________________________________ R-SIG-Finance at stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-finance -- Subscriber-posting only. -- If you want to post, subscribe first.