price break down

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

price break down

Gary Wessle-2
  Hi
   
  Doing calculations on time series data “US bonds” where the price is presented say 11328 to mean 113 28/32, it seams to me that converting the rational would be 0.875 and the round-off error would be expected to cause problems in doing calculations on such numbers, how one could avoid or minimize such a problem?
  or handle bond-kind-of-price in general?  Thanks
   
   
 
               
---------------------------------


        [[alternative HTML version deleted]]


_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
Reply | Threaded
Open this post in threaded view
|

Re: price break down

Spencer Graves
          Nearly all computations in R are to double precision, and I would not
worry about round-off error in this case.

          Far more important is an issue you have not mentioned:  What do you
think about doing essentially all your computations on log price and log
returns?  I recommend this for two reasons:  First, log prices and log
returns tend more nearly normally distributed than the raw data and
unlogged returns.  Second, the logarithms tend to be more tractible
mathematically.  For example, extrapolation from a model fit to prices
in dollars could give you negative prices, i.e., you would have to pay
someone to take your bonds.  By contrast, negative log prices just means
that the price is less than one dollar (or one Swiss Franc or whatever
currency you are using).  If you honestly can be required to pay someone
to take your bonds, then you don't want logarithms;  otherwise, I think
  you do.

          hope this helps.
          spencer graves

Fred J. wrote:

>   Hi
>    
>   Doing calculations on time series data “US bonds” where the
price is presented say 11328 to mean 113 28/32, it seams to me
that converting the rational would be 0.875 and the round-off
error would be expected to cause problems in doing calculations
on such numbers, how one could avoid or minimize such a problem?

>   or handle bond-kind-of-price in general?  Thanks
>    
>    
>  
>
> ---------------------------------
>
>
> [[alternative HTML version deleted]]
>
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> [hidden email] mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-finance

_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
Reply | Threaded
Open this post in threaded view
|

Re: price break down

davidr-2
In reply to this post by Gary Wessle-2
All bond prices are exact in binary arithmetic, so no round-off.

(I may have missed part of this conversation while on vacation.)

David L. Reiner
 

> -----Original Message-----
> From: [hidden email] [mailto:r-sig-finance-
> [hidden email]] On Behalf Of Spencer Graves
> Sent: Monday, March 20, 2006 8:12 PM
> To: Fred J.
> Cc: [hidden email]
> Subject: Re: [R-sig-finance] price break down
>
>  Nearly all computations in R are to double precision, and I
would
> not
> worry about round-off error in this case.
>
>  Far more important is an issue you have not mentioned:  What
do
> you
> think about doing essentially all your computations on log price and
log
> returns?  I recommend this for two reasons:  First, log prices and log
> returns tend more nearly normally distributed than the raw data and
> unlogged returns.  Second, the logarithms tend to be more tractible
> mathematically.  For example, extrapolation from a model fit to prices
> in dollars could give you negative prices, i.e., you would have to pay
> someone to take your bonds.  By contrast, negative log prices just
means
> that the price is less than one dollar (or one Swiss Franc or whatever
> currency you are using).  If you honestly can be required to pay
someone
> to take your bonds, then you don't want logarithms;  otherwise, I
think

>   you do.
>
>  hope this helps.
>  spencer graves
>
> Fred J. wrote:
>
> >   Hi
> >
> >   Doing calculations on time series data "US bonds" where the
> price is presented say 11328 to mean 113 28/32, it seams to me
> that converting the rational would be 0.875 and the round-off
> error would be expected to cause problems in doing calculations
> on such numbers, how one could avoid or minimize such a problem?
> >   or handle bond-kind-of-price in general?  Thanks
> >
> >
> >
> >
> > ---------------------------------
> >
> >
> > [[alternative HTML version deleted]]
> >
> >
> >
> >
------------------------------------------------------------------------
> >
> > _______________________________________________
> > [hidden email] mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-sig-finance
>
> _______________________________________________
> [hidden email] mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-finance

_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance