wild.binary.segmentation {HDcpDetect} | R Documentation |
wild.binary.segmentation
Description
Detects change points in time series data using the wild binary segmentation algorithm from Fryzlewicz (2014).
Usage
wild.binary.segmentation(data_M,minsize=15,num_intervals=1250,M_threshold=0.05)
Arguments
data_M |
An nxp matrix representing a times series of length n with p dimensions. |
minsize |
The minimum interval length. |
num_intervals |
The number of random intervals to be generated and tested for change points. |
M_threshold |
Value used as a threshold to estimate temporal dependence by determining how small of a standardized difference is indistinguishable from zero. |
Details
Increasing the minimum interval length will generally reduce type I error while increasing type II error.
Value
The returned value is a list of the estimated change point locations.
Author(s)
Jun Li, Jeffrey Okamoto, and Natasha Stewart
References
Li, J., Li, L., Xu, M., Zhong, P (2018). Change Point Detection in the Mean of High-Dimensional Time Series Data under Dependence. Manuscript. Fryzlewicz, P. (2014). Wild Binary Segmentation for Multiple Change-point Detection. The Annals of Statistics.
Examples
library(HDcpDetect)
HAPT2 <- as.matrix(HAPT[1:35,])
wild.binary.segmentation(HAPT2)