HDcpDetect: Detect Change Points in Means of High Dimensional Data

Objective: Implement new methods for detecting change points in high-dimensional time series data. These new methods can be applied to non-Gaussian data, account for spatial and temporal dependence, and detect a wide variety of change-point configurations, including changes near the boundary and changes in close proximity. Additionally, this package helps address the “small n, large p” problem, which occurs in many research contexts. This problem arises when a dataset contains changes that are visually evident but do not rise to the level of statistical significance due to the small number of observations and large number of parameters. The problem is overcome by treating the dimensions as a whole and scaling the test statistics only by its standard deviation, rather than scaling each dimension individually. Due to the computational complexity of the functions, the package runs best on datasets with a relatively large number of attributes but no more than a few hundred observations.

Getting started

Package details

AuthorJeffrey Okamoto [aut], Natasha Stewart [aut], Dr. Jun Li [aut]
MaintainerNatasha Stewart <[email protected]>
Package repositoryView on CRAN
Installation Install the latest version of this package by entering the following in R:

Try the HDcpDetect package in your browser

Any scripts or data that you put into this service are public.

HDcpDetect documentation built on Sept. 29, 2018, 1:04 a.m.