Module parametric 

Parametric discount curves: a smooth shape with a handful of parameters.

The pillar families elsewhere in this set fit exactly — one parameter per quoted instrument, and every quote reprices. That is the right tool for a handful of on-the-runs and the wrong one for a government curve, where fifty bonds and four parameters wanted means an exact fit interpolates the noise.

These forms are the other tool. They are smoothers: they will not reprice their inputs, and the fit that produces them reports by how much (see curves::fit). Reach for the calibration ladder when exactness matters.

Every parameter here is a free one, so a curve carrying one of these families reports its sensitivities on that family’s own axis, and a fit can seed them through the substituted walk — which is why the fit’s Jacobian is taken rather than differenced.

Structs

ExponentialSplines

An exponential-splines discount curve (the Vasicek-Fong family).

ExponentialSplinesView

ExponentialSplines lifted into one pricing mode.

NelsonSiegel

A Nelson-Siegel discount curve.

NelsonSiegelView

NelsonSiegel lifted into one pricing mode.

Svensson

A Svensson discount curve — Nelson-Siegel with a second hump.

SvenssonView

Svensson lifted into one pricing mode.

Constants

DECAY_MAX 🔒

See DECAY_MIN.

DECAY_MIN 🔒

The range a decay time constant is allowed to take, in years.

Functions

decay_of 🔒

The decay a stored unconstrained parameter stands for, in mode R.

decay_param 🔒

The unconstrained parameter a decay is stored as. The inverse of decay_of, and the reason a decay outside the range is refused rather than silently clamped: a caller asking for one has said something this form cannot represent.