Struct CusumTest 

pub struct CusumTest<A: Float + Send + Sync> { /* private fields */ }

Two-sided CUSUM control chart.

Uses the reference sample to fix the in-control mean and standard deviation, then accumulates g⁺ = max(0, g⁺ + (x - μ - k)) and g⁻ = max(0, g⁻ - (x - μ + k)) with the classic tabular design k = 0.5σ (half the shift to detect) and h = 5σ (which gives an in-control average run length of roughly 465 for that k). Unlike Page-Hinkley the reference level is fixed rather than running, which makes CUSUM sensitive to slow, sustained drift that a running mean would absorb.

Implementations

impl<A: Float + Send + Sync> CusumTest<A>

pub fn new(sensitivity: f64, significance_level: f64) -> Result<Self, String>

Creates a CUSUM detector. sensitivity scales the decision interval h, so a larger sensitivity fires sooner.

Trait Implementations

impl<A: Float + Default + Clone + Send + Sync + Sum> StatisticalTest<A> for CusumTest<A>

fn test_for_drift( &mut self, reference: &[A], current: &[A], ) -> Result<DriftTestResult<A>, String>

Performs the statistical test for drift

fn update_parameters(&mut self, performance_feedback: A) -> Result<(), String>

Updates test parameters based on historical performance

fn reset(&mut self)

Resets the test state