SNSeg: SNSeg: An R Package for Time Series Segmentation via...

In SNSeg: Self-Normalization(SN) Based Change-Point Estimation for Time Series

SNSeg: An R Package for Time Series Segmentation via Self-Normalization (SN)

Description

The SNSeg package provides three functions for multiple change point estimation using SN-based algorithms: SNSeg_Uni, SNSeg_Multi and SNSeg_HD. Three critical value tables (critical_values_single, critical_values_multi and critical_values_HD) were attached. Functions MAR, MAR_Variance and MAR_MTS_Covariance can be utilized to generate time series data that are used for the functions SNSeg_Uni, SNSeg_Multi and SNSeg_HD. S3 methods plot(), print() and summary() are available for class "SNSeg_Uni", "SNSeg_Multi" and "SNSeh_HD" objects. The function max_SNsweep enables users to compute the SN test statistic and make the segmentation plot for these statistics. The function SNSeh_estimate allows users to compute parameter estimates of each segment that is separated by estimated change-points.

SNSeg_Uni

SNSeg_Uni provides SN-based change point estimates for a univariate time series based on changes in a single parameter or multiple parameters.

For the parameters of the SN test, the function SNSeg_Uni offers mean, variance, acf, bivariate correlation and numeric quantiles as available options. It also allows users to enter their own defined function as the input parameter. Besides, users can use a composite set of parameters including one or more from the mean, variance, acf or numeric quantiles quantile. To visualize the estimated change points, users can set "plot_SN = TRUE" and "est_cp_loc = TRUE" to generate the time series segmentation plot. The output comprises of the parameter(s), the window size, and the estimated change point locations. The function returns an S3 object of class "SNSeg_Uni", which can be applied to S3 methods plot(), print() and summary().

SNSeg_Multi

SNSeg_Multi provides SN-based change point estimates for multivariate time series based on changes in multivariate means or covariance matrix. The "plot_SN = TRUE" option allows users to plot each individual time series and the estimated change=points. The function returns an S3 object of class "SNSeg_Multi", which can be applied to S3 methods plot(), print() and summary().

SNSeg_HD

SNSeg_HD provides SN-based change point estimates for a high-dimensional time series based on changes in high-dimensional means. The "plot_SN = TRUE" option allows users to plot each individual time series and the estimated change=points. The input argument "n_plot" enables users to plot the first "n_plot" number of time series. The function returns an S3 object of class "SNSeg_HD", which can be applied to S3 methods plot(), print() and summary().

max_SNsweep

max_SNsweep provides SN based test statistic of each time point and generates a plot for these statistics and the estimated change-points.

SNSeg_estimate

SNSeg_estimate computes the parameter estimates of each segment separated by the estimated change-points.

critical values table

The package SNSeg provides three critical values table.

Table critical_values_single tabulates critical values of SN-based change point estimates based on the change in a single parameter.

Table critical_values_multi tabulates critical values of SN-based change point estimates based on changes in multiple parameters.

Table critical_values_HD tabulates critical values of of SN-based change point estimates based on changes in high-dimensional means.

SNSeg documentation built on June 22, 2024, 10:50 a.m.