You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Add Nelson-Siegel and Smoothing Spline Yield Curve fitting #70
A good first step would be to make a struct that holds the data for the Nelson-Siegel parametrisation and methods for computing discount and zero rate correctly. Add in some tests. You can get inspired by the Smith-Wilson implementation. The open a draft pr.
Tried to implement and added NelsonSiegelSvensson.jl and SmoothingSplines.jl whose approach might be debatable but please advise how these can be uploaded/pulled for a review.
I have seen there are two spline methods in bootstrapping, do we still need smoothing splines? In this reference https://www.bis.org/publ/bppdf/bispap25.pdf it seems to talk about cubic splines, so smoothing splines here would mean cubic splines, or splines with even higher degrees? Thanks.
My understanding is that "smoothing splines" refers to fitting (minimizing error) of a starting spline model, whereas the existing methods fit simple splines through bootstrapped points (for the given points, the model error is zero).
Review update against FinanceModels v7.0.0 (6cb71714), 2026-10-11.
Partially fulfilled; narrow the remaining feature. Nelson–Siegel and NSS are implemented with reference tests. Spline least-squares fitting, including B-spline interpolation models and fitting an existing knot grid, also exists.
A dedicated roughness-penalized smoothing-spline API is not present. Higher polynomial degree or a smooth interpolant alone does not mean noisy observations are being smoothed, and ordinary least squares is distinct from adding a roughness penalty.
See spline construction and knot-curve fitting. Rename/scope this issue around penalized smoothing, knot selection, and the intended loss/penalty if that capability is still wanted. Medium-priority feature, not a missing NS/NSS implementation.
Activity
alecloudenback commentedon Feb 20, 2022
Related article: https://www.soa.org/sections/financial-reporting/financial-reporting-newsletter/2022/february/fr-2022-02-perelman/
leeyuntien commentedon Mar 7, 2022
Would like to contribute to this issue. Please advise next steps. Thanks!
kasperrisager commentedon Mar 7, 2022
A good first step would be to make a struct that holds the data for the Nelson-Siegel parametrisation and methods for computing discount and zero rate correctly. Add in some tests. You can get inspired by the Smith-Wilson implementation. The open a draft pr.
leeyuntien commentedon Mar 31, 2022
Tried to implement and added NelsonSiegelSvensson.jl and SmoothingSplines.jl whose approach might be debatable but please advise how these can be uploaded/pulled for a review.
alecloudenback commentedon Apr 2, 2022
Awesome! Check out this article about creating a pull request: https://www.digitalocean.com/community/tutorials/how-to-create-a-pull-request-on-github
A PR will allow the automated tests/checks to occur and makes it easy to review the changes.
alecloudenback commentedon Jun 16, 2022
Two of three will be completed in #116
leeyuntien commentedon Jun 22, 2022
I have seen there are two spline methods in bootstrapping, do we still need smoothing splines? In this reference https://www.bis.org/publ/bppdf/bispap25.pdf it seems to talk about cubic splines, so smoothing splines here would mean cubic splines, or splines with even higher degrees? Thanks.
alecloudenback commentedon Jun 23, 2022
My understanding is that "smoothing splines" refers to fitting (minimizing error) of a starting spline model, whereas the existing methods fit simple splines through bootstrapped points (for the given points, the model error is zero).
alecloudenback commentedon Jun 26, 2022
I was looking at DataInterpolations.jl and think that maybe it has smoothing spline fitting built in?
leeyuntien commentedon Jun 27, 2022
I think you are right, and there are examples showing how to call the BSpline functions. https://htmlpreview.github.io/?https://github.com/PumasAI/DataInterpolations.jl/blob/v2.0.0/example/DataInterpolations.html.
alecloudenback commentedon Oct 11, 2026
Review update against FinanceModels v7.0.0 (
6cb71714), 2026-10-11.Partially fulfilled; narrow the remaining feature. Nelson–Siegel and NSS are implemented with reference tests. Spline least-squares fitting, including B-spline interpolation models and fitting an existing knot grid, also exists.
A dedicated roughness-penalized smoothing-spline API is not present. Higher polynomial degree or a smooth interpolant alone does not mean noisy observations are being smoothed, and ordinary least squares is distinct from adding a roughness penalty.
See spline construction and knot-curve fitting. Rename/scope this issue around penalized smoothing, knot selection, and the intended loss/penalty if that capability is still wanted. Medium-priority feature, not a missing NS/NSS implementation.