Short communication: Detecting possibly frequent change-points: wild binary segmentation 2 and steepest-drop model selection

Lund, Robert; Shi, Xueheng

doi:10.1007/s42952-020-00081-6

Cited by 5 publications

(5 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This is in contrast to Fryzlewicz (2018), whose methodology estimates at least 10 changepoints, and Baranowski et al (2019) who find five changepoints in the series. We believe that the most likely explanation for this is the presence of significant autocorrelation within the series, which may cause the methods to overfit the number of changepoints, as noted in Lund and Shi (2020). To illustrate this point, in Figure 5, we plot the estimated autocorrelation for the mean removed Newham series, at lags 1, 2, and 3.…”

Section: Data Applicationsmentioning

confidence: 99%

“…One approach for dealing with autocorrelation when testing for changes in mean is to increase the threshold above which changes are detected; see Lavielle (1999), for example. However, it is difficult to systematically choose the threshold without performing some sort of preestimation of the autocovariance, which is highly challenging in the presence of mean changes (Lund & Shi, 2020). If one simply raises the threshold, then there will be a trade‐off between a decreased false positive rate and a decreased true positive rate.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Detecting changes in mean in the presence of time‐varying autocovariance

McGonigle

Killick

Nunes

2021

Stat

View full text Add to dashboard Cite

There has been much attention in recent years to the problem of detecting mean changes in a piecewise constant time series. Often, methods assume that the noise can be taken to be independent, identically distributed (IID), which in practice may not be a reasonable assumption. There is comparatively little work studying the problem of mean changepoint detection in time series with nontrivial autocovariance structure. In this article, we propose a likelihood‐based method using wavelets to detect changes in mean in time series that exhibit time‐varying autocovariance. Our proposed technique is shown to work well for time series with a variety of error structures via a simulation study, and we demonstrate its effectiveness on two data examples arising in economics.

show abstract

Section: Data Applicationsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Detecting changes in mean in the presence of time‐varying autocovariance

McGonigle

Killick

Nunes

2021

Stat

View full text Add to dashboard Cite

show abstract

“…As previously discussed, most changepoint techniques mistakenly flag changepoints when underlying positive dependence is ignored. For example, [26] argues that shifts identified in the London house price series of [27] may be more attributable to the positive correlations in the series than to actual mean shifts. CUSUM based techniques are known to degrade with positive correlation [12].…”

Section: Changepoints In Ar(p) Seriesmentioning

confidence: 99%

Autocovariance estimation in the presence of changepoints

Gallagher

Killick

Lund

et al. 2022

J. Korean Stat. Soc.

Self Cite

View full text Add to dashboard Cite

This article studies estimation of a stationary autocovariance structure in the presence of an unknown number of mean shifts. Here, a Yule-Walker moment estimator for the autoregressive parameters in a dependent time series contaminated by mean shift changepoints is proposed and studied. The estimator is based on first order differences of the series and is proven consistent and asymptotically normal when the number of changepoints m and the series length N satisfy m/N → 0 as N → ∞.

show abstract

“…Any ARMA(p, q) error model can be fitted to the centered series. A genetic algorithm (GA) can be used to estimate the optimal penalized likelihood over all changepoint configurations under a variety of penalization schemes [Lund and Shi, 2020]. Unfortunately, the GA may take significant computing time to find the optimum: there are 2 N −1 different multiple changepoint configurations in a series of length N to search over.…”

Section: Model and Estimation Approachesmentioning

confidence: 99%

“…6.1 Changepoints in AR(p Series Some techniques can mistakenly flag changepoints when underlying positive dependence is ignored. For example, Lund and Shi [2020] argues that the London house price series shifts identified may be more attributable to the positive correlations in the series than to actual mean shifts. CUSUM based techniques are severely affected by positive correlation Shi et al [2021].…”

Section: Applicationsmentioning

confidence: 99%