Detect stochastic change points in noisy biometric signals

Detects abrupt changes in noisy biological time series using a dependency-light stochastic rolling-window approximation. The score combines adjacent-window changes in mean and variance with a robust adaptive threshold.

Usage

detect_gazepoint_doubly_stochastic_changepoints(
  dat,
  signal_col,
  time_col = "CNT",
  group_cols = NULL,
  window_seconds = 10,
  step_seconds = 2,
  threshold_mad_multiplier = 6,
  min_distance_s = 5
)

Arguments

dat

A data frame.

signal_col

Numeric signal column.

time_col

Numeric time column.

group_cols

Optional grouping columns.

window_seconds

Window length in seconds.

step_seconds

Step size in seconds.

threshold_mad_multiplier

Robust threshold multiplier.

min_distance_s

Minimum distance between detected change points.

Value

A list with overview, score_table, changepoints, and settings.

Details

This is not a full reproduction of any specific doubly stochastic model. It is a transparent approximation for QC and exploratory segmentation.