Non-Parametric Multiple Change-Point Analysis of Multivariate Data

Implements various procedures for finding multiple change-points. Two methods make use of dynamic programming and pruning, with no distributional assumptions other than the existence of certain absolute moments in one method. Hierarchical and exact search methods are included. All methods return the set of estimated change- points as well as other summary information.


News

ecp 3.1.0

Changes made since previous version

  • Fixed bugs and updated pruning step.

ecp 3.0.0

Changes made since previous version

  • Added a new function ks.cp3o. This is a new way to perform change point analysis. It uses dynamic programming and pruning in order to obtain change point estimates. It uses the Kolmogorov-Smirnov statistic as goodness-of-fit measure.

ecp 2.0.0

Changes made since previous version

  • Added a new function e.cp3o. This is a new way to perform change point analysis. It uses dynamic programming and pruning in order to obtain change point estimates. It uses the E-statistic as goodness-of-fit measure.

ecp 1.6.2

Changes made since previous version

  • Updated citation information and added vignette.

ecp 1.6.1

Changes made since previous version

  • Changed name of list element returned by e.agglo. Instead of opt it is now estimates. This is so that its behaviour more closely resembles that of e.divisive.

ecp 1.6.0

Changes made since previous version

  • Removed the potential for the permutation test to terminate early. This was causing memory issues, and will be brought back once these issues have been resolved.

ecp 1.5.6

Changes made since previous version

  • Fixed issue with examples taking too long to run when installing package.

ecp 1.5.5

Change made since previous version

  • The DJIA dataset now includes the univariate market index, as well as the multivariate time series created by the DOW30 companies. Kraft is not included in the multivariate series.

  • The list elements in the DJIA dataset are now named; "date", "index", and "market". The "index" component used to be named "returns". The "market" component contains the multivariate time series.

ecp 1.5.4

Changes made since previous version

  • Some of the components of the list returned by both methods have been renamed.

  • The member argument for e.agglo now has a default value.

  • Fixed issue with the e.agglo penalty argument.

  • The penalty argument for e.agglo now accepts a function that depends only on the change point locations.

  • Both the e.agglo and e.divisive methods now also return the estimated membership vector.

  • The e.divisive method now also returns a vector containing the number of permutations performed at each stage of the sequential testing procedure.

  • The STPP dataset has been removed from the package.

ecp 1.5.3

Changes made since previous version

  • The agglomerative procedure now includes an argument for a penalty term.

  • New permutation test implementation. The new version is based on the method of Gandy (2009), and R now is the maximum number of permutations to perform.

  • Added 2 additional arguments to the e.divisive() method. These arguments are necessary to implement the updated permutation test.

  • Function argument names now adhere to standard R naming conventions.

  • Names of returned list components now adhere to standard R naming convention.

  • In the list returned by the e.divisive() method, the list component has been renamed estimates.

  • Updated documentation.

ecp 1.5.2

Changes made since previous version

  • Reordered the argument list for the e.agglo() method. The call is now e.agglo(data, member, alpha=1).

  • In the list returned by the e.agglo() method, the gof component has been renamed fit.

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.

install.packages("ecp")

3.1.0 by Wenyu Zhang, a year ago


Browse source code at https://github.com/cran/ecp


Authors: Nicholas A. James , Wenyu Zhang and David S. Matteson


Documentation:   PDF Manual  


Task views: Time Series Analysis


GPL (>= 2) license


Depends on Rcpp

Suggests mvtnorm, MASS, combinat, R.rsp

Linking to Rcpp


Imported by PCRedux.


See at CRAN