Idris A. Eckley's research works | Lancaster Unive...

... The nonparametric Pettitt test is one of the most widely applied, and it allows the user to determine the timing of the change point and its significance (Pettitt, 1979). The performance of the Pettitt test -and other similar approaches, including pruned exact linear time (PELT; Killick et al., 2012), binary segmentation (Scott and Knott, 1974), Bayesian analysis (Erdman and Emerson, 2008), wild binary segmentation (WBS; Fryzlewicz et al., 2014), nonparametric PELT (Haynes et al., 2017), and the Mann-Whitney test (Mann and Whitney, 1947) -for the estimation of abrupt changes in the mean, variance, or median of a time series was tested using both simulated and historical data with known change points (Ryberg et al., 2019). Although these methods offer potential benefits (such as the detection of multiple change points), the comparison found that the Pettitt test delivered the best combination of change point detection and minimization of false positive results. ...