[R-SIG-Finance] Need help please

Wed Jan 27 18:36:52 CET 2010

Hi Bogaso,

I am not a VaR expert at all, but I would say this:
1. Not enough data to compute anything with any sensible level of  
statistical significance (probably already very weak for volatility,  
but most probably very very very weak for VaR estimation)
2. I do not know this "very notorious" time series personally  
(probably my own limitation), but it looks pretty odd to me, basically  
you have 7 different prices repeated many times giving you 6 returns -  
one of which is almost 50% - and many nil returns...
3. VaR using the simple log-normal model of asset prices (I guess you  
mean VaR(99%)=average - 2.33*stdev) is vastly documented to be  
insufficient for financial assets like equity, so most probably not  
for whatever this thing is.

So I would personally simply say that it is not possible with the data  
at hand. Better to accept you cannot than use inappropriate "models"  
just to get some number out. But that's a personal opinion. Or there  
is another complex way (probably with a set of strong assumptions on  
the variable) which is beyond my knowledge

& as Ivan pointed out, this does not even enter the debate whether VaR  
is a sensible risk measure (coherence, best outcome in a bad day,  
etc...). I may be wrong of course, but I would be interested to have  
other list member opinion on this.

Sorry, this is most probably not the hoped/expected feedback.

Rgds,
Julen

On Jan 27, 2010, at 5:44 PM, Zhang, Ivan wrote:

> Hi Bogaso,
>
> This may not necessarily help you get an answer, but perhaps would  
> steer
> you in another direction:
>
> If the series doesn't look continuous you may potentially be able to
> pick a quantile that would make this measure not "coherent" which
> basically invalidates the use of VaR as a measure of risk in this  
> case.
>
> For more information on Coherent risk measures, see below link or  
> Google
> "coherent risk measure"
>
> http://www.math.ethz.ch/~delbaen/ftp/preprints/CoherentMF.pdf
>
>
> Hope this helps,
>
>
> -Ivan Zhang
>
> -----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 Bogaso
> Sent: Tuesday, January 26, 2010 3:25 AM
> To: r-sig-finance at stat.math.ethz.ch
> Subject: [R-SIG-Finance] Need help please
>
>
> Dear folks,
>
> I got a very notorious weekly price series where price seldom changes
> like :
>
> 6-Jan-92	4.38
> 13-Jan-92	4.38
> 20-Jan-92	4.38
> 27-Jan-92	4.38
> 3-Feb-92	4.38
> 10-Feb-92	4.38
> 17-Feb-92	4.38
> 24-Feb-92	4.38
> 2-Mar-92	4.38
> 9-Mar-92	4.38
> 16-Mar-92	4.38
> 23-Mar-92	4.38
> 30-Mar-92	4.38
> 6-Apr-92	4.38
> 13-Apr-92	4.38
> 20-Apr-92	6.56
> 27-Apr-92	6.56
> 4-May-92	6.56
> 11-May-92	6.56
> 18-May-92	6.56
> 25-May-92	6.56
> 1-Jun-92	6.56
> 8-Jun-92	6.63
> 15-Jun-92	6.63
> 22-Jun-92	6.63
> 29-Jun-92	6.63
> 6-Jul-92	6.63
> 13-Jul-92	6.63
> 20-Jul-92	6.99
> 27-Jul-92	6.99
> 3-Aug-92	6.99
> 10-Aug-92	6.99
> 17-Aug-92	6.99
> 24-Aug-92	6.99
> 31-Aug-92	6.99
> 7-Sep-92	6.99
> 14-Sep-92	6.99
> 21-Sep-92	6.99
> 28-Sep-92	6.99
> 5-Oct-92	6.99
> 12-Oct-92	6.99
> 19-Oct-92	6.99
> 26-Oct-92	6.99
> 2-Nov-92	6.99
> 9-Nov-92	6.99
> 16-Nov-92	6.99
> 23-Nov-92	6.99
> 30-Nov-92	6.99
> 7-Dec-92	6.99
> 14-Dec-92	6.99
> 21-Dec-92	6.99
> 28-Dec-92	6.99
> 4-Jan-93	6.99
> 11-Jan-93	6.99
> 18-Jan-93	6.99
> 25-Jan-93	6.99
> 1-Feb-93	6.99
> 8-Feb-93	6.99
> 15-Feb-93	6.99
> 22-Feb-93	6.99
> 1-Mar-93	6.99
> 8-Mar-93	6.99
> 15-Mar-93	6.99
> 22-Mar-93	6.56
> 29-Mar-93	6.56
> 5-Apr-93	6.56
> 12-Apr-93	6.56
> 19-Apr-93	6.56
> 26-Apr-93	6.56
> 3-May-93	6.56
> 10-May-93	6.63
> 17-May-93	6.63
> 24-May-93	6.63
> 31-May-93	6.63
> 7-Jun-93	6.63
> 14-Jun-93	6.63
> 21-Jun-93	6.99
> 28-Jun-93	6.99
> 5-Jul-93	6.99
> 12-Jul-93	6.99
> 19-Jul-93	6.99
> 26-Jul-93	6.99
> 2-Aug-93	6.99
> 9-Aug-93	6.99
> 16-Aug-93	6.99
> 23-Aug-93	6.99
>
> I have a mandate to calculate VaR on that price data, probably in
> Parametric
> way. My question is can I apply standard way which we generally use  
> like
> log-normally distributed price, to calculate VaR here? Or some other
> modeling approach needs to be taken care? Can anyone please provide me
> any
> references over net, how to handle this type of scenario?
>
> Your help will be highly appreciated.
>
> Thanks,
> -- 
> View this message in context:
> http://n4.nabble.com/Need-help-please-tp1290132p1290132.html
> Sent from the Rmetrics mailing list archive at Nabble.com.
>
> _______________________________________________
> 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.
> -- Also note that this is not the r-help list where general R  
> questions
> should go.
>
> _______________________________________________
> 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.
> -- Also note that this is not the r-help list where general R  
> questions should go.
>