Optimal detection of changepoints with a linear computational cost

We consider the problem of detecting multiple changepoints in large data sets. Our focus is on applications where the number of changepoints will increase as we collect more data: for example in genetics as we analyse larger regions of the genome, or in finance as we observe time-series over longer periods. We consider the common approach of detecting changepoints through minimising a cost function over possible numbers and locations of changepoints. This includes several established procedures for detecting changing points, such as penalised likelihood and minimum description length. We introduce a new ∗R. Killick is Senior Research Associate, Department of Mathematics & Statistics, Lancaster University, Lancaster, UK (E-mail: r.killick@lancs.ac.uk). P. Fearnhead is Professor, Department of Mathematics & Statistics, Lancaster University, Lancaster, UK (E-mail: p.fearnhead@lancs.ac.uk). I.A. Eckley is Senior Lecturer, Department of Mathematics & Statistics, Lancaster University, Lancaster, UK (E-mail: i.eckley@lancs.ac.uk). The authors are grateful to Richard Davis and Alice Cleynen for providing the Auto-PARM and PDPA software respectively. Part of this research was conducted whilst R. Killick was a jointly funded Engineering and Physical Sciences Research Council (EPSRC) / Shell Research Ltd graduate student at Lancaster University. Both I.A. Eckley and R. Killick also gratefully acknowledge the financial support of the EPSRC grant number EP/I016368/1. 1 ar X iv :1 10 1. 14 38 v3 [ st at .M E ] 9 O ct 2 01 2 method for finding the minimum of such cost functions and hence the optimal number and location of changepoints that has a computational cost which, under mild conditions, is linear in the number of observations. This compares favourably with existing methods for the same problem whose computational cost can be quadratic or even cubic. In simulation studies we show that our new method can be orders of magnitude faster than these alternative exact methods. We also compare with the Binary Segmentation algorithm for identifying changepoints, showing that the exactness of our approach can lead to substantial improvements in the accuracy of the inferred segmentation of the data.

[1]  O. L. Smith,et al.  ON THE CUMULANTS OF RENEWAL PROCESSES , 1959 .

[2]  Stuart E. Dreyfus,et al.  Applied Dynamic Programming , 1965 .

[3]  H. Akaike.  A new look at the statistical model identification , 1974 .

[4]  A. Scott,et al.  A Cluster Analysis Method for Grouping Means in the Analysis of Variance , 1974 .

[5]  John H. Holland,et al.  Adaptation in Natural and Artificial Systems: An Introductory Analysis with Applications to Biology, Control, and Artificial Intelligence , 1992 .

[6]  M. Srivastava,et al.  On Tests for Detecting Change in Mean , 1975 .

[7]  D. Hsu.  Tests for Variance Shift at an Unknown Time Point , 1977 .

[8]  G. Schwarz.  Estimating the Dimension of a Model , 1978 .

[9]  Yi-Ching Yao.  Estimation of a Noisy Discrete-Time Step Function: Bayes and Empirical Bayes Approaches , 1984 .

[10]  I E Auger,et al.  Algorithms for the optimal identification of segment neighborhoods. , 1989, Bulletin of mathematical biology.

[11]  L. Horváth,et al.  The Maximum Likelihood Method for Testing Changes in the Parameters of Normal Observations , 1993 .

[12]  G. C. Tiao,et al.  Use of Cumulative Sums of Squares for Retrospective Detection of Changes of Variance , 1994 .

[13]  C. Inclan,et al.  Volatility in Emerging Stock Markets , 1997, Journal of Financial and Quantitative Analysis.

[14]  Jianfeng Yao,et al.  On the Underfitting and Overfitting Sets of Models Chosen by Order Selection Criteria , 1999 .

[15]  H. Müller,et al.  Multiple changepoint fitting via quasilikelihood, with application to DNA sequence segmentation , 2000 .

[16]  Arjun K. Gupta,et al.  Parametric Statistical Change Point Analysis , 2000 .

[17]  E. Ghysels,et al.  Detecting Multiple Breaks in Financial Market Volatility Dynamics , 2002 .

[18]  M. Wigler,et al.  Circular binary segmentation for the analysis of array-based DNA copy number data. , 2004, Biostatistics.

[19]  Franck Picard,et al.  A statistical approach for array CGH data analysis , 2005, BMC Bioinformatics.

[20]  Jeffrey D. Scargle,et al.  An algorithm for optimal partitioning of data on an interval , 2003, IEEE Signal Processing Letters.

[21]  Q. Shao,et al.  On discriminating between long-range dependence and changes in mean , 2006, math/0607803.

[22]  Richard A. Davis,et al.  Structural Break Estimation for Nonstationary Time Series Models , 2006 .

[23]  David O Siegmund,et al.  A Modified Bayes Information Criterion with Applications to the Analysis of Comparative Genomic Hybridization Data , 2007, Biometrics.

[24]  J. Steinebach,et al.  On the detection of changes in autoregressive time series I. Asymptotics , 2007 .

[25]  P. Massart,et al.  Minimal Penalties for Gaussian Model Selection , 2007 .

[26]  Edit Gombay.  Change detection in autoregressive time series , 2008 .

[27]  Servane Gey,et al.  Using CART to Detect Multiple Change Points in the Mean for Large Sample , 2008 .

[28]  Paul Fearnhead,et al.  Bayesian Analysis of Isochores , 2009 .

[29]  Guillem Rigaill,et al.  Pruned dynamic programming for optimal multiple change-point detection , 2010 .

[30]  Z. Harchaoui,et al.  Multiple Change-Point Estimation With a Total Variation Penalty , 2010 .

[31]  P. Fearnhead,et al.  Analysis of changepoint models. , 2011 .

[32]  Stéphane Robin,et al.  Joint segmentation, calling, and normalization of multiple CGH profiles. , 2011, Biostatistics.