[R-SIG-Finance] regarding bootstrapping... REVISITED

[Brian G. Peterson]

On Tuesday 10 October 2006 00:25, gyadav at ccilindia.co.in wrote:
> i am trying to build a spot yield curve for fixed income market
> specifically bonds. i was told by my contacts that this can be done
> best by bootstrapping.


On Tuesday 10 October 2006 01:44, gyadav at ccilindia.co.in wrote:
> I went through the thread(
> https://stat.ethz.ch/pipermail/r-sig-finance/2006q1/000682.html which
> concerns with swaps). Yeah it is correct that i would like to quote
> both David and Krishna that the curve interpolation may vary
> considerably (for e.g. any polynomial/parametric fit is very different
> from and curve fitting whether it is free hand or by NURBS ( complex
> version of Basis Splines ZZZzzz). My problem is that i want to know how
> can i generate spot curve using bootstrap method in R.Further, even if
> you do not have fixed maturity bonds i.e. when you need to create
> fictitious or virtual paper of varied fixed maturities like 1 month, 6
> month, 1 year, 5 year, 10 year ..... so that you can create a spot
> curve from the traded points which may be like as follows.... for e.g.

Gaurav,

I believe we're all saying the same thing.  

David has correctly pointed out that to simply discuss "bootstrapping a 
yield curve" does not necessarily imply the use of a statistical 
"bootstrap" method to build your curve, although it does not necessarily 
rule it out either.  Kris in the earlier thread provided code for an 
interpolation/fit method using the discount rate. Thomas used a different 
method.  Kris also provided reference to several papers that could be 
used to construct other methods, and pointed out that the choice of 
method will change your estimates, possibly significantly. My point was 
that you will need to test any fitting method against the specific 
problem that you have, and that a statistical bootstrap may or may not be 
appropriate to your problem. The input data you have available will help 
you determine the best fitting method to use.

Both Thomas' code and Kris' code look like they will do a credible job of 
fitting a yield curve.  Perhaps you should consider testing those methods 
against your problem, so that you could identify deficiencies that those 
methods may have in your specific implementation.  Then we could discuss 
approaches here that might address the specific deficiencies that you 
identify.

Regards,

   - Brian