HDCD (version 1.0)

single_SBS_calibrate: Generates threshold \(\pi_T\) for Sparsified Binary Segmentation for single change-point detection

Description

R wrapper for function choosing empirical threshold \(\pi_T\) using Monte Carlo simulation for single change-point Sparsified Binary Segmentation. More specifically, the function returns the empirical upper tol quantile of CUSUMs over \(p\) time series, each of length \(n\), based on \(N\) number of runs.

Usage

single_SBS_calibrate(
  n,
  p,
  N = 100,
  tol = 1/100,
  rescale_variance = TRUE,
  debug = FALSE
)

Value

Threshold

Arguments

n

Number of observations

p

Number time series

N

Number of Monte Carlo samples used

tol

False positive probability tolerance

rescale_variance

If TRUE, each row of the data is rescaled by a MAD estimate

debug

If TRUE, diagnostic prints are provided during execution

Examples

library(HDCD)
n = 50
p = 50
set.seed(101)

# Simulate threshold
pi_T_squared = single_SBS_calibrate(n=n,p=p,N=100, tol=1/100, rescale_variance = TRUE)
pi_T_squared


# Generating data
X = matrix(rnorm(n*p), ncol = n, nrow=p)
# Adding a single sparse change-point:
X[1:5, 26:n] = X[1:5, 26:n] +1

# Run SBS
res = single_SBS(X,threshold=sqrt(pi_T_squared),rescale_variance=TRUE)
res$pos