| Type: | Package |
| Title: | Nonparametric Multiple Change Point Detection Using Wild Binary Segmentation |
| Version: | 1.0 |
| Author: | Gordon J. Ross [aut, cre] |
| Maintainer: | Gordon J. Ross <gordon.ross@ed.ac.uk> |
| Description: | Implements nonparametric multiple change-point detection for univariate sequences using Wild Binary Segmentation, as described in Ross (2026) "Nonparametric Detection of Multiple Location-Scale Change Points via Wild Binary Segmentation" <doi:10.48550/arXiv.2107.01742>. The package provides Mann–Whitney, Mood, Lepage, Cramér–von Mises, modified Baumgartner, standardised Zhang Z_C, and standardised Anderson–Darling rank-based statistics, together with method-specific thresholds for controlling the probability of incorrectly detecting a change point in a homogeneous sequence. |
| Depends: | R (≥ 4.0.0) |
| Imports: | digest, Rcpp |
| LinkingTo: | Rcpp |
| License: | GPL-3 |
| Encoding: | UTF-8 |
| NeedsCompilation: | yes |
| Config/npwbs/Development-Status: | M=1000 calibrations imported |
| Packaged: | 2026-10-06 04:01:50 UTC; rosss |
| Repository: | CRAN |
| Date/Publication: | 2026-10-06 04:20:02 UTC |
Nonparametric detection of multiple change points using Wild Binary Segmentation
Description
Returns the estimated number and locations of the change points in a sequence of univariate observations. For full details of how this procedure works, please see G. J. Ross (2026) - "Nonparametric Detection of Multiple Location-Scale Change Points via Wild Binary Segmentation" at https://arxiv.org/abs/2107.01742
Usage
detectChanges(y, alpha=0.05, prune=TRUE, M=1000, d=2,
displayOutput=FALSE, method="lepage", breakTies=TRUE,
combination="sum")
Arguments
y |
A non-empty finite numeric vector to test for change points. Matrices and other dimensioned inputs are rejected. |
alpha |
Required Type I error rate. Lepage with |
prune |
Whether to prune potential excess change points via post-processing. Most likely should be left as TRUE. |
M |
Number of subsequences to sample during WBS and pruning. The default is |
d |
Endpoint trimming within each sampled interval. The existing methods, including Anderson–Darling, use |
displayOutput |
A single non-missing logical. If TRUE, prints information throughout the recursive search for change points. |
method |
Rank statistic: exactly one of |
breakTies |
If TRUE, tied inputs receive one random strict ordering before detection; if FALSE, ties are an error. |
combination |
For |
Details
Verified M = 1000 tables are used by default. Exact frozen M = 10000 tables remain available for all original methods. Unsupported method/combination/M/alpha/d tuples fail rather than reuse a mismatched table. The sampled interval minimum length is fixed at 10. Thresholds for Lepage, Mann–Whitney, Mood, Cramér–von Mises, and modified Baumgartner are provided through segment length n = 10000; longer inputs retain the package's existing behaviour of reusing the n = 10000 threshold with one warning. Zhang and Anderson–Darling support lengths only through n = 3000.
The "zhang" method is the exact mean/standard-deviation-standardised Zhang Z_C detector used in the paper. It retains d = 4, alpha = 0.05, strict threshold exceedance, and registered length-specific thresholds for M = 1000 and M = 10000. Exact moments through n = 1000 are bundled. Sequence lengths n = 1001 through 3000 require one prior call to download_zhang_moments; an invalid installed artifact can be replaced with download_zhang_moments(overwrite = TRUE). Lengths above 3000 are unsupported. Anderson–Darling supports only M = 1000, alpha = 0.05, d = 2, and lengths through 3000.
If prune = TRUE, a post-processing pruning step retests merged neighbouring segments and removes changepoints that are no longer supported.
Value
A vector containing the detected changepoint locations. A returned value k denotes a split between observations k and k + 1, i.e. the change occurs after observation k.
Author(s)
Gordon J. Ross gordon.ross@ed.ac.uk
Examples
set.seed(100)
y <- c(rnorm(30,0,1),rnorm(30,3,1), rnorm(30,0,1),rnorm(30,0,3))
detectChanges(y)
Install the extended exact Zhang moment table
Description
Downloads and verifies the fixed exact-moment artifact required by the standardised Zhang detector for sequence lengths above 1000.
Usage
download_zhang_moments(overwrite = FALSE)
Arguments
overwrite |
If |
Details
Exact moments through sequence length 1000 are bundled with npwbs.
The method = "zhang" detector requires this one-time download for
lengths 1001 through 3000. The artifact is installed in the package-specific
user data directory returned by tools::R_user_dir("npwbs", "data").
Use overwrite = TRUE to reinstall an invalid or incompatible artifact.
Sequence lengths above 3000 are unsupported.
Value
The installed file path, returned invisibly.