penCPD: Penalized change point detection

In Hamedhm/SmoothWin: Soft Windowing on Linear Regression

Description

The implementation of penalized change point detection algorithm on a signal.

Usage

  penCPD(
    x                  ,
    plot = FALSE       ,
    threshold = 0.01   ,
    method = "enet"    ,
    criteria = "AICc"  ,
    ...
  )

Arguments

x	Vector. Vector of the data
plot	Logical flag. Set to true returns some plots
threshold	Single positive value. Threshold to filter the change points. The effect sizes below the threshold would be pushed to zero
method	Case sensitive string. Choose between 'enet' and 'alasso' for lasso and adaptive-lasso change point detection. Defaults is 'enet'.
criteria	Case sensitive string. Choose between 'Cp', 'AICc', 'BIC' and 'GCV' for the model selection criteria in penalized algorithm. The default is 'AICc'.
...	parameters that are passed to the plot() function

Value

a vector of change point indices. For example 5 means the fifth value in x

Author(s)

Hamed Haselimashhadi <[e-mail]>

References

On change point detection using the fused lasso method, CR Rojas, B Wahlberg - arXiv preprint arXiv:1401.5408, 2014

Examples

  set.seed(123456)
  n   = 100
  x   = c(rnorm(n,0,.1),rnorm(n,1,.1))
  CPs = penCPD(x = x,plot = TRUE,method = 'alasso')
  plot(x)
  abline(v=CPs)

Hamedhm/SmoothWin documentation built on May 18, 2019, 10:13 a.m.