If you run the example on the help page (example() won't work, but just
copy/paste it) it certainly looks like the intention is that VaR is a left-tail
thing, so usually negative. E.g.
> sum(actual < VaR)
[1] 74
> summary(VaR)
Index VaR
Min. :1991-02-27 00:00:00 Min. :-0.04919
1st Qu.:1992-08-19 18:00:00 1st Qu.:-0.02765
Median :1994-02-10 12:00:00 Median :-0.02368
Mean :1994-02-12 12:01:36 Mean :-0.02492
3rd Qu.:1995-08-08 06:00:00 3rd Qu.:-0.02109
Max. :1997-01-30 00:00:00 Max. :-0.01594
this is somewhat contrary to conventional definition where VaR is an _upper_
quantile in a _loss_ distribution, which of course differs from the _return_
distribution by sign and a multiplication with the invested amount. I gather,
however, that the conventions are not too solid and may e.g. vary between
textbooks.
-pd
> On 29 Jul 2017, at 16:11 , T.Riedle <tr...@kent.ac.uk> wrote:
>
> Dear all,
>
> I want to backtest my Value at Risk output using the VaRTest() function in
> the rugarch package. I do not understand if the numeric vector of VaR which
> needs to be calculated is in negative or positive terms. Usually VaR is
> expressed in positive terms.
>
>
>
> Do I have to use positive values for VaR in the VaRTest() formula?
>
>
>
> Thanks for your help.
>
>
>
>
>
> [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
--
Peter Dalgaard, Professor,
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Office: A 4.23
Email: pd....@cbs.dk Priv: pda...@gmail.com
______________________________________________
R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
