VARDetect (version 0.1.8)

first.step.detect: First step rolling window function

Description

Rolling window scheme function for the first step

Usage

first.step.detect(
  data,
  h,
  step.size = NULL,
  lambda,
  mu,
  alpha_L = 0.25,
  skip = 3,
  lambda.1.seq = NULL,
  mu.1.seq = NULL,
  cv = FALSE,
  nfold = NULL,
  verbose = FALSE
)

Value

A vector which includes all candidate change points selected by rolling window

Arguments

data

the whole data matrix

h

window size

step.size

rolling step size, default is NULL. If Null, the step size is 1/4 of the window size

lambda

a 2-d vector of tuning parameters for sparse components, available when cv is FALSE

mu

a 2-d vector of tuning parameters for low rank components, available when cv is FALSE

alpha_L

a numeric value, indicates the size of constraint space of low rank component

skip

the number of observations we should skip near the boundaries, default is 3

lambda.1.seq

the sequence of sparse tuning parameter to the left segment, only available when cv is TRUE

mu.1.seq

the sequence of low rank tuning, only available for cv is TRUE

cv

a boolean argument, indicates whether use cross validation or not

nfold

a positive integer, indicates the number of folds of cross validation

verbose

if TRUE, then all information for current stage are printed