Error when using a Cubic interpolator for PiecewiseYieldCurve
Locked
Error when using a Cubic interpolator for PiecewiseYieldCurve
 
|
I'm just starting out with QuantlLib and started going through the
examples . In an attempt to get smoother forwards, I would like to use
a cubic (or log-cubic, or monotonic) interpolator instead of the
log-linear interpolator. So I modified the example code "swapvaluation.cpp"
by
replacing
boost::shared_ptr<YieldTermStructure> depoFutSwapTermStructure( new PiecewiseYieldCurve< Discount, LogLinear>( settlementDate, depoFutSwapInstruments, termStructureDayCounter, tolerance)); with boost::shared_ptr<YieldTermStructure> depoFutSwapTermStructure( new PiecewiseYieldCurve< Discount, Cubic >( settlementDate, depoFutSwapInstruments, termStructureDayCounter, tolerance)); I get the following error message: 1st iteration: failed at 10th instrument, maturity December 20th, 2006, reference date September 22nd, 2004: root not bracketed: f[2.22045e-016,0.932437] -> [-3.323214e+018,-3.911669e+000] The full output is: Today: Monday, September 20th, 2004 Settlement date: Wednesday, September 22nd, 2004 ==================================================================== 5-year market swap-rate = 4.43 % ==================================================================== 5-years swap paying 4.00 % term structure | net present value | fair spread | fair fixed rate | -------------------------------------------------------------------- depo-swap | 19065.88 | -0.42 % | 4.43 % | 1st iteration: failed at 10th instrument, maturity December 20th, 2006, reference date September 22nd, 2004: root not bracketed: f[2.22045e-016,0.932437] -> [-3.323214e+018,-3.911669e+000] Is there something additional I must do when using an interpolator other than LogLinear? Thanks, -- Robert Philipp Synapse Financial Engineering 703.623.4063 (mobile) 703.573.0119 (fax) [hidden email] www.synapsefe.com ------------------------------------------------------------------------------ The Palm PDK Hot Apps Program offers developers who use the Plug-In Development Kit to bring their C/C++ apps to Palm for a share of $1 Million in cash or HP Products. Visit us here for more details: http://p.sf.net/sfu/dev2dev-palm _______________________________________________ QuantLib-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-users |
Re: Error when using a Cubic interpolator for PiecewiseYieldCurve
|
|
Robert Philipp schrieb:
> I'm just starting out with QuantlLib and started going through the > examples . In an attempt to get smoother forwards, I would like to use > a cubic (or log-cubic, or monotonic) interpolator instead of the > log-linear interpolator. So I modified the example code > "swapvaluation.cpp" by replacing > > boost::shared_ptr<YieldTermStructure> depoFutSwapTermStructure( > new PiecewiseYieldCurve< Discount, *LogLinear*>( > settlementDate, > depoFutSwapInstruments, > termStructureDayCounter, > tolerance)); > > with > boost::shared_ptr<YieldTermStructure> depoFutSwapTermStructure( > new PiecewiseYieldCurve< Discount, *Cubic *>( > settlementDate, > depoFutSwapInstruments, > termStructureDayCounter, > tolerance)); > > I get the following error message: > 1st iteration: failed at 10th instrument, maturity December > 20th, 2006, reference date September 22nd, 2004: > root not bracketed: f[2.22045e-016,0.932437] -> > [-3.323214e+018,-3.911669e+000] assertions to detect ill-posed problems and throws an error if the code is being misused. Take a look at lines 195 : QL_REQUIRE(fxMin_*fxMax_ < 0.0, "root not bracketed: f[" << xMin_ << "," << xMax_ << "] -> [" << std::scientific << fxMin_ << "," << fxMax_ << "]"); For testing you can comment it out, and calculate the results. It might be that the assertion is being too restrictive. Do you mind to supply the data that you want to bootstrapp? ( Instrument, Maturity,value,...) HTH -Kim > > The full output is: > Today: Monday, September 20th, 2004 > Settlement date: Wednesday, September 22nd, 2004 > > ==================================================================== > 5-year market swap-rate = 4.43 % > > ==================================================================== > 5-years swap paying 4.00 % > term structure | net present value | fair spread | fair fixed > rate | > > -------------------------------------------------------------------- > depo-swap | 19065.88 | -0.42 % | > 4.43 % | > 1st iteration: failed at 10th instrument, maturity December > 20th, 2006, reference date September 22nd, 2004: > root not bracketed: f[2.22045e-016,0.932437] -> > [-3.323214e+018,-3.911669e+000] > > Is there something additional I must do when using an interpolator > other than LogLinear? > > Thanks, > -- > Robert Philipp > Synapse Financial Engineering > 703.623.4063 (mobile) > 703.573.0119 (fax) > > [hidden email] > www.synapsefe.com > ------------------------------------------------------------------------ > > ------------------------------------------------------------------------------ > The Palm PDK Hot Apps Program offers developers who use the > Plug-In Development Kit to bring their C/C++ apps to Palm for a share > of $1 Million in cash or HP Products. Visit us here for more details: > http://p.sf.net/sfu/dev2dev-palm > ------------------------------------------------------------------------ > > _______________________________________________ > QuantLib-users mailing list > [hidden email] > https://lists.sourceforge.net/lists/listinfo/quantlib-users > ------------------------------------------------------------------------------ The Palm PDK Hot Apps Program offers developers who use the Plug-In Development Kit to bring their C/C++ apps to Palm for a share of $1 Million in cash or HP Products. Visit us here for more details: http://p.sf.net/sfu/dev2dev-palm _______________________________________________ QuantLib-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-users |
Re: Error when using a Cubic interpolator for PiecewiseYieldCurve
|
|
The data the poster uses is the data in the example code.
I am struggling with the same issue. Other people have suggested looking at the fitted bond curve example, but that is somewhat different. I just want to calc forward swap rates. Commenting out the check seems a little dangerous. Any help with this would be appreciated. The documentation is a little thin. If it helps, I am messing around with different cubic methods. Here's how you spec the interpolation with a little more control. Check the documentation here: http://quantlib.org/reference/class_quant_lib_1_1_cubic_interpolation.html Cubic cube(CubicInterpolation::FritschButland, true, CubicInterpolation::Lagrange, 0,CubicInterpolation::Lagrange, 0); boost::shared_ptr<YieldTermStructure> termStructure( new PiecewiseYieldCurve<Discount,Cubic>( settlementDate, depoFutSwapInstruments, termStructureDayCounter, tolerance, cube)); ------------------------------------------------------------------------------ The Palm PDK Hot Apps Program offers developers who use the Plug-In Development Kit to bring their C/C++ apps to Palm for a share of $1 Million in cash or HP Products. Visit us here for more details: http://p.sf.net/sfu/dev2dev-palm _______________________________________________ QuantLib-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-users |
Re: Error when using a Cubic interpolator for PiecewiseYieldCurve
|
|
MonkeyMan <feersum_monkey <at> yahoo.com> writes:
For what it's worth, here is the data I'm using (not the example data): Today: Sunday, January 4th, 2004 Settlement date: Wednesday, January 7th, 2004 deposit: sw 0.0108375 deposit: 2w 0.0109125 deposit: 1m 0.0112 deposit: 2m 0.0113 deposit: 3m 0.0115 deposit: 6m 0.012275 deposit: 9m 0.0135 deposit: 1y 0.0151 future: 2004-03-17 98.765 future: 2004-06-16 98.525 future: 2004-09-15 98.145 future: 2004-12-15 97.695 future: 2005-03-16 97.26 future: 2005-06-15 96.855 future: 2005-09-21 96.51 future: 2005-12-21 96.21 swap: 10y 4.758 swap: 12y 4.973 swap: 15y 5.21 swap: 2y 2.245 swap: 20y 5.405 swap: 3y 2.875 swap: 30y 5.484 swap: 4y 3.36 swap: 5y 3.755 swap: 6y 4.045 swap: 7y 4.268 swap: 8y 4.46 swap: 9y 4.622 Then I blow up on the first instrument when I try to price a 1yf2y swap with this error: 1st iteration: failed at 1st instrument, maturity January 14th, 2004, reference date January 7th, 2004: root not bracketed: f[2.22045e-16,1] ->[1.083750e-02,1.083750e-02] For this instance I used the ConvexMonotone: Real quadraticity = 0.3; Real monotonicity = 0.7; bool forcePositive = true; ConvexMonotone monotone = ConvexMonotone( quadraticity, monotonicity, forcePositive); boost::shared_ptr<YieldTermStructure> termStructure( new PiecewiseYieldCurve<Discount,ConvexMonotone>( settlementDate, depoFutSwapInstruments, termStructureDayCounter, tolerance, monotone)); Note that my swap rates are pushed into the RateHelper out of order. This is the order they come out of the database. I assume that doesn't matter. Is there some better way to do this? The parameters for the ConvexMonotone are the defaults. I have tried using the Hagan parameters, but that didn't help. What do other people do to bootstrap their curves? Am I missing something basic? ------------------------------------------------------------------------------ The Palm PDK Hot Apps Program offers developers who use the Plug-In Development Kit to bring their C/C++ apps to Palm for a share of $1 Million in cash or HP Products. Visit us here for more details: http://p.sf.net/sfu/dev2dev-palm _______________________________________________ QuantLib-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-users |
Re: Error when using a Cubic interpolator for PiecewiseYieldCurve
|
|
In reply to this post by MonkeyMan-3
MonkeyMan <feersum_monkey <at> yahoo.com> writes:
MonkeyMan <feersum_monkey <at> yahoo.com> writes: For what it's worth, here is the data I'm using (not the example data): Today: Sunday, January 4th, 2004 Settlement date: Wednesday, January 7th, 2004 deposit: sw 0.0108375 deposit: 2w 0.0109125 deposit: 1m 0.0112 deposit: 2m 0.0113 deposit: 3m 0.0115 deposit: 6m 0.012275 deposit: 9m 0.0135 deposit: 1y 0.0151 future: 2004-03-17 98.765 future: 2004-06-16 98.525 future: 2004-09-15 98.145 future: 2004-12-15 97.695 future: 2005-03-16 97.26 future: 2005-06-15 96.855 future: 2005-09-21 96.51 future: 2005-12-21 96.21 swap: 10y 4.758 swap: 12y 4.973 swap: 15y 5.21 swap: 2y 2.245 swap: 20y 5.405 swap: 3y 2.875 swap: 30y 5.484 swap: 4y 3.36 swap: 5y 3.755 swap: 6y 4.045 swap: 7y 4.268 swap: 8y 4.46 swap: 9y 4.622 Then I blow up on the first instrument when I try to price a 1yf2y swap with this error: 1st iteration: failed at 1st instrument, maturity January 14th, 2004, reference date January 7th, 2004: root not bracketed: f[2.22045e-16,1] ->[1.083750e-02,1.083750e-02] For this instance I used the ConvexMonotone: Real quadraticity = 0.3; Real monotonicity = 0.7; bool forcePositive = true; ConvexMonotone monotone = ConvexMonotone( quadraticity, monotonicity, forcePositive); boost::shared_ptr<YieldTermStructure> termStructure( new PiecewiseYieldCurve<Discount,ConvexMonotone>( settlementDate, depoFutSwapInstruments, termStructureDayCounter, tolerance, monotone)); Note that my swap rates are pushed into the RateHelper out of order. This is the order they come out of the database. I assume that doesn't matter. Is there some better way to do this? The parameters for the ConvexMonotone are the defaults. I have tried using the Hagan parameters, but that didn't help. What do other people do to bootstrap their curves? Am I missing something basic? ------------------------------------------------------------------------------ The Palm PDK Hot Apps Program offers developers who use the Plug-In Development Kit to bring their C/C++ apps to Palm for a share of $1 Million in cash or HP Products. Visit us here for more details: http://p.sf.net/sfu/dev2dev-palm _______________________________________________ QuantLib-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-users |
|
|
In reply to this post by Robert Philipp
you need to pair the interpolation method correctly with the curve output...
if using discount factors: MonotonicLogCubicNaturalSpline or LogLinear should work if using zero rates: try MonotonicCubicNaturalSpline or CubicNaturalSpline the reason you are getting an error is because the bootstrapping for the entire curve cant be completed with the interpolation method you selected (cubic on discount factors). even if it did work, the swap rates you would have received would have been off-market. On Thu, Jul 29, 2010 at 12:31 PM, Robert Philipp <[hidden email]> wrote:
------------------------------------------------------------------------------ The Palm PDK Hot Apps Program offers developers who use the Plug-In Development Kit to bring their C/C++ apps to Palm for a share of $1 Million in cash or HP Products. Visit us here for more details: http://p.sf.net/sfu/dev2dev-palm _______________________________________________ QuantLib-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-users |
Re: Error when using a Cubic interpolator for PiecewiseYieldCurve
|
|
In reply to this post by Robert Philipp
Thanks for the advice. That seems to work for me for some test dates. Here is
the explicit call I used to create the MonotonicLogCubicNaturalSpline for anyone else following this: //MonotonicLogCubicNaturalSpline LogCubic mlcns = LogCubic( CubicInterpolation::Spline, true, CubicInterpolation::SecondDerivative, 0, CubicInterpolation::SecondDerivative, 0) ; boost::shared_ptr<YieldTermStructure> termStructure( new PiecewiseYieldCurve<Discount,LogCubic>( settlementDate, depoFutSwapInstruments, termStructureDayCounter, tolerance, mlcns)); Also, apologies for the multiple posts before. Didn't think it went through... Monkey ------------------------------------------------------------------------------ The Palm PDK Hot Apps Program offers developers who use the Plug-In Development Kit to bring their C/C++ apps to Palm for a share of $1 Million in cash or HP Products. Visit us here for more details: http://p.sf.net/sfu/dev2dev-palm _______________________________________________ QuantLib-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-users |
«
Return to quantlib-users
|
1 view|%1 views
| Free forum by Nabble | Edit this page |