Analytical approach to changepoint detection in Laplacian noise

The paper presents an analytical method using the Bayesian inference framework for the identification of time-series discontinuities, i.e. changepoints, in impulsive Laplacian noise. Exact expressions for the posterior density of the changepoint positions and the associated Bayesian model evidence are given for DC step changes. The performance of the analytical approach is compared to that predicted by a Gaussian assumption to the noise statistics and Markov chain Monte Carlo methods.