Bayesian change-point modeling with segmented ARMA model - Fig 6

(Top) Segmented signal with the true change-point locations. (Middle) Posterior probabilities of occurrence of change-points. (Bottom) Estimated change-point locations (posterior probability greater than 0.5). The middle plot shows the location of peaks in the probability profile closely follows the true change-points locations but in some positions with low posterior probability. Using a threshold in posterior probability 0.5, we identify 17 change-points out of 19 which match the locations of the true change-points.