Risky Measures of Risk: Error Analysis of Numerical Differentiation
80 Pages Posted: 9 Jan 2007
Date Written: June 22, 2005
We review and detail the causes of errors in numerical differentiation, including roundoff error, convexity error, cancellation error and correlated errors. We discuss methods for improving accuracy, including step size selection and smoothing techniques, as well as a number of approaches specific to the types of computations being done.
Keywords: greeks, delta, gamma, theta, differentiation, numerical, numerical differentiation, error, error control, roundoff, convexity, cancellation, smoothing, Hermite, filtering, convolution, Fourier, Gaussian, quadrature, Complex
JEL Classification: C63, G13
Suggested Citation: Suggested Citation