Numerical approximation of probability mass functions via the inverse discrete Fourier transform

DOI10.1007/S11009-013-9366-3MaRDI QIDQ2513667zbMATH OpenOpenAlexFDO

Authors Richard L. Warr

Publication date 28 January 2015

Published in Methodology and Computing in Applied Probability (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1212.6546

zbMATH Keywords

characteristic functiondiscrete Fourier transformfast Fourier transformfirst passage distributionsemi-Markov process

Mathematics Subject Classification ID

Stationary stochastic processes (60G10) Markov renewal processes, semi-Markov processes (60K15) Numerical methods for discrete and fast Fourier transforms (65T50)

Abstract: First passage distributions of semi-Markov processes are of interest in fields such as reliability, survival analysis, and many others. The problem of finding or computing first passage distributions is, in general, quite challenging. We take the approach of using characteristic functions (or Fourier transforms) and inverting them, to numerically calculate the first passage distribution. Numerical inversion of characteristic functions can be numerically unstable for a general probability measure, however, we show for lattice distributions they can be quickly calculated using the inverse discrete Fourier transform. Using the fast Fourier transform algorithm these computations can be extremely fast. In addition to the speed of this approach, we are able to prove a few useful bounds for the numerical inversion error of the characteristic functions. These error bounds rely on the existence of a first or second moment of the distribution, or on an eventual monotonicity condition. We demonstrate these techniques in an example and include R-code.

Review text

Let \(T\) be a random variable with distribution \(F\). The characteristic function and Fourier transform of \(T\) are defined as \[ \varphi_T(s)=\int_{-\infty}^{\infty}e^{ist}dF(t) \] and \[ \hat{f}(\omega)=\int_{-\infty}^{\infty}e^{-2\pi i \omega}dF(t), \] respectively. The author suggests using the inverse fast Fourier transform for the inversion of the characteristic function to probability mass functions. To make it the author derives error bounds for lattice distributions. For example, the next statement is proved. { Lemma 1.} For a nonnegative lattice random variable \(T\), defined on the support \(n\Delta t\), \(n=0,1,\dots,\infty\), the pointwise error of the forward discrete Fourier transform is less or equal to \(\operatorname{P}(T\geq N\Delta t)\).

Recommendations

• Error Bounds for Numerical Inversion of a Probability Characteristic Function
• Inverting analytic characteristic functions and financial applications
• The Fourier-series method for inverting transforms of probability distributions
• Numerical inversion of a characteristic function
• Approximation of probabilistic Laplace transforms and their inverses

Cites work

• Application of high order numerical quadratures to numerical inversion of the Laplace transform
• Error Bounds for Numerical Inversion of a Probability Characteristic Function
• First-passage-time moments of Markov processes
• Flowgraph Models for Multistate Time‐to‐Event Data
• scientific article; zbMATH DE number 2063761 (Why is no real title available?)
• scientific article; zbMATH DE number 771803 (Why is no real title available?)
• scientific article; zbMATH DE number 3237318 (Why is no real title available?)
• Markov Renewal Processes with Finitely Many States
• Multi-precision Laplace transform inversion
• Semi-Markov chains and hidden semi-Markov models toward applications. Their use in reliability and DNA analysis.
• The first-passage times of phase semi-Markov processes
• The Fourier-series method for inverting transforms of probability distributions

Cited in

(10)

• The Fourier-series method for inverting transforms of probability distributions
• Likelihood computation in the normal-gamma stochastic frontier model
• Using semi-Markov chains to solve semi-Markov processes
• Error bounds for cumulative distribution functions of convolutions via the discrete Fourier transform
• Error Bounds for Numerical Inversion of a Probability Characteristic Function
• On the discretisation error in the computation of the empirical characteristic function
• Improved FFT approximations of probability functions based on modified quadrature rules
• Bayesian nonparametric estimation of first passage distributions in semi-Markov processes
• Bootstrapping through discrete convolutional methods
• A fiducial-based confidence interval for the linear combination of multinomial probabilities

Describes a project that uses

Uses Software

• R

This page was built for publication: Numerical approximation of probability mass functions via the inverse discrete Fourier transform

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2513667)