Change-point Method Research Articles

Detection of spatiotemporal changepoints: a generalised additive model approach

The detection of changepoints in spatio-temporal datasets has been receiving increased focus in recent years and is utilised in a wide range of fields. With temporal data observed at different spatial locations, the current approach is typically to use univariate changepoint methods in a marginal sense with the detected changepoint being representative of a single location only. We present a spatio-temporal changepoint method that utilises a generalised additive model (GAM) dependent on the 2D spatial location and the observation time to account for the underlying spatio-temporal process. We use the full likelihood of the GAM in conjunction with the pruned linear exact time (PELT) changepoint search algorithm to detect multiple changepoints across spatial locations in a computationally efficient manner. When compared to a univariate marginal approach our method is shown to perform more efficiently in simulation studies at detecting true changepoints and demonstrates less evidence of overfitting. Furthermore, as the approach explicitly models spatio-temporal dependencies between spatial locations, any changepoints detected are common across the locations. We demonstrate an application of the method to an air quality dataset covering the COVID-19 lockdown in the United Kingdom.

Precipitation forecast estimation applying the change point method and ARIMA

Precipitation forecast estimation applying the change point method and ARIMA

Research on Newborns Forecasting in China-based on the Three-Child Policy

The essay uses data from 1921-2021 to forecast the impact of three-child policy on births based on the National Bureau of Statistics of China. By using different methods (Arima, Holt, Change Point, Functional Data Analysis), this paper draws a conclusion that the number of newborns in China is likely to be an upward trend after the publication of three-child policy. To be precise, the result from the Arima model has no reference value, while the results from Holt’s method and the change point method indicate that the number of babies born in China will rise. Referring to the change point analysis, it represents a downward trend if there are not any adjustments. However, after combining the data with world birth tendency, the result shows that the number of newborns in China is growing in the next following years. These results represent that the publication of China’s three child policy has a positive effect on the figure for new babies born in China.

A Change-Point Method to Detect Meaningful Change in Return-to-Sport Progression in Athletes.

To explore how the change-point method can be used to analyze complex longitudinal data and detect when meaningful changes (change points) have occurred during rehabilitation. This design is a prospective single-case observational study of a football player in a professional club who sustained an acute lower-limb muscle injury during high-speed running in training. The rehabilitation program was entirely completed in the football club under the supervision of the club's medical team. Four wellness metrics and 5 running-performance metrics were collected before the injury and until the player returned to play. Data were collected over 130days. In the univariate analysis, the change points for stress, sleep, mood, and soreness were located on days 30, 47, 50, and 50, respectively. The change points for total distance, acceleration, maximum speed, deceleration, and high-speed running were located on days 32, 34, 37, 41, and 41, respectively. The multivariate analysis resulted in a single change point for the wellness metrics and running-performance metrics, on days 50 and 67, respectively. The univariate approach provided information regarding the sequence and time point of the change points. The multivariate approach provided a common change point for multiple metrics, information that would benefit clinicians to have a broad overview of the changes in the rehabilitation process. Clinicians may consider the change-point method to integrate and visualize data from multiple sources to evaluate athletes' progression along the return-to-sport continuum.

Elastic functional changepoint detection of climate impacts from localized sources

AbstractDetecting changepoints in functional data has become an important problem as interest in monitoring of climate phenomenon has increased, where the data is functional in nature. The observed data often contains both amplitude (‐axis) and phase (‐axis) variability. If not accounted for properly, true changepoints may be undetected, and the estimated underlying mean change functions will be incorrect. In this article, an elastic functional changepoint method is developed which properly accounts for these types of variability. The method can detect amplitude and phase changepoints which current methods in the literature do not, as they focus solely on the amplitude changepoint. This method can easily be implemented using the functions directly or can be computed via functional principal component analysis to ease the computational burden. We apply the method and its nonelastic competitors to both simulated data and observed data to show its efficiency in handling data with phase variation with both amplitude and phase changepoints. We use the method to evaluate potential changes in stratospheric temperature due to the eruption of Mt. Pinatubo in the Philippines in June 1991. Using an epidemic changepoint model, we find evidence of a increase in stratospheric temperature during a period that contains the immediate aftermath of Mt. Pinatubo, with most detected changepoints occurring in the tropics as expected.

Analysis of time-series monitoring data for early detection of CO2 leakage to shallow subsurface environments

Time-series data of environmental parameters have been gathered in numerous artificial CO2 release experiments performed globally; however, their usability must be re-evaluated because they are generally utilized as auxiliary data, such as model construction or system operation checks. In this study, we propose a process for determining the environmental changepoint (EnCP) using continuous data for the reliable fast detection of unexpected CO2 intrusion into a shallow subsurface. From a controlled CO2 gas release test, continuous data on electrical conductivity, pH, and partial pressure of CO2 (pCO2) in the saturated zone, as well as CO2 concentration in the vadose zone were obtained and analyzed using four methods (simple excess, simple moving average, offline, and online change-point detection) to determine the environmental change-point caused by CO2 appearance. We found that changes caused by CO2 gas injection occurred rapidly, allowing the EnCP detected in the time-series analysis to identify changes more efficiently than the occasional sampling, which was performed at least once a day. It was confirmed that pCO2 had the highest sensitivity among the saturated zone indicators, and the online change-point method was more successful than other methods in determining EnCP effectively. Furthermore, the determined EnCPs at each monitoring point, showing the front boundary of the injected CO2, allowed for the comprehensive understanding of CO2 behavior in both the saturated and vadose zones. This may be used to aid in the design and operation of a real-time CO2 leakage surveillance network system for the early detection of CO2 intrusion into shallow subsurface environments.

Choice of degree of Bernstein polynomial model

Methods based on moments and modes for choosing Bernstein polynomial model degree are proposed. Simulation showed that the mean integrated square errors of the maximum approximate Bernstein likelihood estimates of density using degrees selected by the proposed methods and the change-point method are smaller than that of the kernel density but either closer to or smaller than that of the parametric maximum likelihood estimator.

A change point method to inform athlete progression in the return to sport progress

Introduction: Return-to-sport (RTS) decision-making is often challenging, as rehabilitation is complex and non-linear. With advances in technology, an increasing volume of data is being collected at multiple time points during rehabilitation to track the progression. Computer-based analytical methods, such as change point (CP) detection, can leverage the data collected longitudinally to inform clinicians of the time points when there were meaningful changes. Here, we exemplify how longitudinal data in wellness and rehabilitation running performance data could be analysed with the CP approach.

Optimizing Management Practices to Reduce Sediment Connectivity between Forest Roads and Streams in a Mountainous Watershed

Forest roads often increase runoff and sediment loss, thus greatly impacting hydrological processes in mountainous watersheds. While there has been previous investigation on best management practices (BMPs) to reduce soil erosion from forest roads, few studies have attempted to optimize BMPs based on how much they can decrease sediment connectivity between forest roads and streams. To close this gap in knowledge, we analyzed the spatial relationship between forest roads and streams, presented the spatial distribution of sediment connectivity by integrating the forest roads into the calculation of the index of connectivity (IC), determined how sediment connectivity would respond to additional BMPs through simulating scenarios, and used these data to optimize the BMPs so they would intercept the greatest sediment loads. We found that forest roads and streams in the Xiangchagou watershed in the Dabie Mountain area of China tend to occur within 180 m of each other; however, within the same buffer zones, streams are more often accompanied by forest roads. IC was greatest near road–stream crossings but smaller near streams and forest roads, and it tended to decrease as the buffer distance increased. Furthermore, we found that sediment connectivity was decreased through running a variety of scenarios that used sediment basin and riparian buffers as BMPs between forest roads and streams. Specifically, within this watershed, riparian buffers should be 64 m wide, and there should be 30 sediment basins with a minimum upslope drainage area of 2 ha. At these quantities, the BMPs in this watershed would significantly affect sediment connectivity. By contrast, beyond these thresholds, increasing the width of riparian buffers or the number of sediment basins does not lead to meaningful sediment reductions. In this way, we were able to use the mean change point method to determine the optimal sediment basin quantity (30 with corresponding minimum upslope drainage area of 2 ha) and the optimal riparian buffer width (64 m) for the Xiangchagou watershed. While these results are a first approximation in a novel research area, they can guide forest managers and stakeholders to design and optimize BMPs that control the delivery of eroded sediments associated with forest roads.

Change-Point detection for autocorrelated multivariate Poisson processes

ABSTRACT Statistical process control for count data has attracted increasing attention in recent years. The need for efficient control charts suitable for autocorrelated multivariate count processes is well recognized. However, there is a scarcity of research aiming to take into account the autocorrelation among the multivariate count data. We are motivated to study the Phase I analysis of autocorrelated multivariate Poisson processes to detect and estimate change points in reference datasets. A change-point method is proposed based on the multivariate Poisson INAR(1) model by integrating generalized likelihood ratio tests with the binary segmentation procedure. A diagnostic procedure for pinpointing the location of the change point is also discussed. Our simulation results show that the proposed method has a better performance than the benchmark method, across a range of possible shifts, in the detection effectiveness and diagnostic accuracy. Furthermore, a real example from the manufacturing industry is used to illustrate the implementation steps of the proposed method.

Single Gaussian process method for arbitrary tokamak regimes with a statistical analysis

Gaussian process regression is a Bayesian method for inferring profiles based on input data. The technique is increasing in popularity in the fusion community due to its many advantages over traditional fitting techniques including intrinsic uncertainty quantification and robustness to over-fitting. This work investigates the use of a new method, the change-point method, for handling the varying length scales found in different tokamak regimes. The use of the Student’s t-distribution for the Bayesian likelihood probability is also investigated and shown to be advantageous in providing good fits in profiles with many outliers. To compare different methods, synthetic data generated from analytic profiles is used to create a database enabling a quantitative statistical comparison of which methods perform the best. Using a full Bayesian approach with the change-point method, Matérn kernel for the prior probability, and Student’s t-distribution for the likelihood is shown to give the best results.

Abnormal data cleaning of wind turbine power curve using Bayesian change point-quartile combined algorithm

During the operation of wind turbines, there are a lot of abnormal data collected from wind power data, which cannot be used to truly evaluate the performance and operation state of wind turbines. According to the distribution characteristics of these abnormal data, an outlier delete method using Bayesian change point-quartile combined algorithm is proposed. Firstly, to identify and delete the abnormal data below wind power curve, based on the Bayesian theorem, change point position is identified. Finally, the quartile method is utilized to delete the upper sparse outliers. The proposed method is applied to 10-min field monitoring wind power data, root mean square error (RMSE), program running time and the rejection rate of field wind turbine data are used to measure model accuracy, the results show that compared with the least squares (LS) change point method, the method proposed in this paper is feasible and effective.

On the non-parametric changepoint detection of flow regimes in cyclone Amphan

The Bay of Bengal was witness to a severe cyclone named Amphan during the summer of the year 2020. The National Institute of Ocean Technology (NIOT), INDIA moorings BD08 and BD09 happened to be in the vicinity of the cyclone. The highly instrumented mooring recorded near-surface meteorological parameters like wind speed, sea surface temperature, and near-surface pressure. This article explores the possibility of using a non-parametric algorithm to identify different flow regimes using a one-month long time-series data of the near-surface parameters. The changes in the structure of the time series signal were statistically segmented using an unconstrained non-parametric algorithm. The non-parametric changepoint method was applied to time series of near-surface winds, sea surface temperature, sea level pressure, air temperature and salinity and the segmentations are consistent with visual observations. Identifying different data segments and their simple parameterization is a crucial component and relating them to different flow regimes is useful for the development of parametrization schemes in weather and climate models. The segmentations can considerably simplify the parametrization schemes when expressed as linear functions. Moreover, the usefulness of non-parametric automatic detection of data segments of similar statistical properties shall be more apparent when dealing with relatively long time series data.

Development of a Portable Infrared-Type Noncontact Blood Pressure Measuring Device and Evaluation of Blood Pressure Elevation during Driving

Hypertension has been established as a major risk factor for cardiovascular morbidity and mortality. Therefore, prevention of hypertension is an urgent matter to maintain people’s health and avoid death, and out-of-office blood pressure measurement is said to be an integral part of the diagnosis and management of hypertension. Hypertension not only causes loss of productivity and economic loss but is also a major cause of road accidents. Therefore, it is important to develop an in-vehicle noncontact blood pressure measurement system for drivers. In addition to measurement accuracy, proper detection timing is also important, and there must be no difference between noncontact and contact detection times. In this study, we introduce an infrared cuffless and portable noncontact blood pressure monitoring system, its measurement principle, and performance evaluation. A total of 13 male adults participated in the experiment to evaluate the effect of time lag between the use of the infrared blood pressure monitoring system and the contact blood pressure monitoring system using a driving simulator. The changepoint method was applied to detect the first change point in the blood pressure time series data caused by the unexpected first appearance of the vehicle. The results showed that the detection time of the developed system was about 2.5 s shorter than that of the contact-type continuous blood pressure measurement system, with no significant difference.

Neck Circumference and Risk of Incident Atrial Fibrillation in the Framingham Heart Study.

BackgroundIncreased neck circumference, a proxy for upper‐body subcutaneous fat, is associated with cardiovascular risk and metabolic risk factors, accounting for body mass index (BMI) and waist circumference. The association between neck circumference and incident atrial fibrillation (AF) is unclear. The aim of current study was to evaluate the association between neck circumference and incident AF.Methods and ResultsWe selected participants from the Framingham Heart Study aged ≥55 years without diagnosed AF and with available neck circumference, BMI, and waist circumference measurements. We defined high neck circumference as ≥14 inches in women and ≥17 inches in men on the basis of the Contal and O’Quigley changepoint method. We used Fine‐Gray models to estimate subdistribution hazards ratios (sHRs) for the association between neck circumference and incident AF accounting for the competing risk of death. We adjusted models for clinical risk factors. We then additionally adjusted separately for BMI, waist circumference, and height/weight. The study sample included 4093 participants (mean age 64±7 years, 55% female). During 11.2±5.7 mean years of follow‐up, incident AF occurred in 571 participants. High neck circumference was associated with incident AF (sHR for high versus low: 1.58; 95% CI, 1.32–1.90, P<0.0001). The association remained significant after adjustment for BMI (sHR, 1.51; 95% CI, 1.21–1.89; P=0.0003), waist circumference (sHR, 1.47; 95% CI, 1.18–1.83; P<0.0001), and height/weight (sHR, 1.37; 95% CI, 1.09–1.72; P=0.007).ConclusionsHigh neck circumference was associated with incident AF adjusting for traditional adiposity measures such as BMI and waist circumference.

The effects of terrain factors on the drainage area threshold: comparison of principal component analysis and correlation analysis.

How and to which extent terrain factors affecting the drainage area threshold (DAT) are disputable. This paper uses principal component analysis (PCA) and correlation analysis to study the influence degree of terrain factors on DAT. Firstly, 22 watersheds, locating in the severe soil erosion region (SSER) of Loess Plateau of China, are picked out as the example areas. The purpose of the mean change point method (MCP) to detect the relationship between DAT and gully density (GD) is to get a reasonable DAT. Secondly, nine terrain factors are calculated, and their statistical values are compared and put in the matrix to clear the different effects on DAT. Finally, the effects of statistical eigenvalues of terrain factors on DAT are compared with PCA and the correlation analysis. According to the PCA, the nine terrain factors are summarized into three principal components, which are slope, height variation, and relief factor. By calculating the score weighted by each factor coefficient matrix and eigenvalue, the result states that slope (S), terrain curvatures (K), and surface roughness (SR) are the factors that have great influence on DAT. Meanwhile, the results of correlation analysis indicate that S, SR, and K have exerted a great influence on the DAT, and the significance level was above 0.05. Both the results of PCA and correlation analysis make clear that the slope is the most direct and influential factor affecting DAT, while other factors are more or less related to slope directly and indirectly. The result implies that the vertical variation of terrain has a strong correlation with the slope, and also has a great influence on DAT. This research not only would be of great significance to recognize the mechanism of gully development, but also able to provide a scientific reference for soil and water conservation in the Loess Plateau.

Cross-covariance isolate detect: A new change-point method for estimating dynamic functional connectivity.

Evidence of the non stationary behavior of functional connectivity (FC) networks has been observed in task based functional magnetic resonance imaging (fMRI) experiments and even prominently in resting state fMRI data. This has led to the development of several new statistical methods for estimating this time-varying connectivity, with the majority of the methods utilizing a sliding window approach. While computationally feasible, the sliding window approach has several limitations. In this paper, we circumvent the sliding window, by introducing a statistical method that finds change-points in FC networks where the number and location of change-points are unknown a priori. The new method, called cross-covariance isolate detect (CCID), detects multiple change-points in the second-order (cross-covariance or network) structure of multivariate, possibly high-dimensional time series. CCID allows for change-point detection in the presence of frequent changes of possibly small magnitudes, can assign change-points to one or multiple brain regions, and is computationally fast. In addition, CCID is particularly suited to task based data, where the subject alternates between task and rest, as it firstly attempts isolation of each of the change-points within subintervals, and secondly their detection therein. Furthermore, we also propose a new information criterion for CCID to identify the change-points. We apply CCID to several simulated data sets and to task based and resting state fMRI data and compare it to recent change-point methods. CCID may also be applicable to electroencephalography (EEG), magentoencephalography (MEG) and electrocorticography (ECoG) data. Similar to other biological networks, understanding the complex network organization and functional dynamics of the brain can lead to profound clinical implications. Finally, the R package ccid implementing the method from the paper is available from CRAN.

A parameter‐free exponential control chart based on Kullback‐Leibler information for time‐between‐events monitoring

AbstractThis paper proposes a Kullback‐Leibler information (KLI) control chart for monitoring the occurrence rate of a process that follows an exponential distribution. The chart is designed for detecting a decrease, an increase, or both in the occurrence rate. Due to the skewness of the plotted statistic, the performance of the chart is investigated separately when the occurrence rate is estimated by the maximum likelihood estimator, the uniformly minimum variance unbiased estimator, and the minimum mean squared error estimator. The KLI chart is effective for a wide range of shifts of the occurrence rate. Its performance is slightly behind cumulative sum (CUSUM) and exponentially weighted moving average (EWMA) charts around sizes of shift where these two charts are optimized. It outperforms these two charts elsewhere. The relative mean index (RMI) values show that the KLI chart has better overall performance. A backward empirical sequential test is used for the plotted statistic of the chart. It is shown equivalent to the cumulative sum scheme and the change point method for the generalized likelihood ratio chart in terms of run length. The use of KLI chart does not require any design parameter other than the control limit.

Orographic Influences on the Spatial Distribution of PM2.5 on the Fen-Wei Plain

Topography not only has an effect on the spatial distribution of pollutants by restricting populations and industrial activities, but also affects the formation, transmission, accumulation, diffusion, and sedimentation of PM2.5. Topography is, therefore, an essential variable for understanding the spatial distribution of PM2.5, especially in basin areas. Based on gridded data and both natural and human indexes, this paper examines orographic effects on the distribution of PM2.5 on the Fen-Wei Plain based on the change point method, regression, the geographical detector method, and bivariate spatial autocorrelation. The results indicate that:① The relief amplitude of the plain is generally lower in its central part than at its edge, which is similar to the attitudinal distribution in this region; ② The distribution of PM2.5 is negatively related to relief amplitude, with high concentrations in central area and lower concentrations at the edge; ③ Based on the geographical detector analysis, PM2.5 patterns in this region are shaped by human factors including population, gross domestic product(GDP), and energy consumption, as well as natural factors including meteorological conditions and vegetation; and ④ Topography has a significant effect on both human and natural factors; an increase of relief amplitude is associated with a decrease in population, GDP, and energy consumption based on power function. Accordingly, temperature goes down linearly; precipitation, relative moisture, and wind speed increase linearly; and the vegetation index increases based on a logarithmic function.

How to identify the different phases of stock market bubbles statistically?

Eugene Fama once mentioned in 2016 that people have not come up with ways of identifying bubbles statistically. This paper presents the nonparametric change-point method to identify different stages of stock bubbles, and we derive its asymptotic distribution under the null hypothesis. By simulation, we obtain the corresponding critical value. In the empirical analysis, we employ this test and binary segmentation method to the 1990s Nasdaq bubble and get the same result as Phillips et al. (2011). We also apply this test to the S&P 500 index, the Shanghai stock index, the Nikkei 225 index, the FTSE 100 index, and the CAC 40 index respectively, and successfully identify the bubbles’ different phases in each stock market.