Struct PearsonCorrelation 

pub struct PearsonCorrelation;

Pearson correlation coefficient.

Implementations

impl PearsonCorrelation

pub fn compute(x: &[f64], y: &[f64]) -> Result<f64, MultivariateError>

Compute Pearson correlation between two equal-length slices.

NaN pairs are deleted pairwise — rolling_corr does the same within each window. When either series has zero variance the correlation is undefined and the result is NaN, not 0.0: 0.0 asserts “these series are uncorrelated” where the truth is “there is nothing to correlate”, and a caller acts on the first.

That sentence used to claim rolling_corr shared this convention while rolling_corr folded a NaN into a sliding accumulator and nulled the rest of its partition. A doc comment asserting that two implementations agree is worth a test that they do; see chronix/tests/analytics_null_semantics.rs.

Errors

MultivariateError::DimensionMismatch on unequal lengths, and MultivariateError::InsufficientData with fewer than two usable pairs.

Trait Implementations

impl CorrelationMethod for PearsonCorrelation

fn compute(&self, x: &[f64], y: &[f64]) -> Result<f64, MultivariateError>

Compute the correlation coefficient between two equal-length slices.

fn supports_incremental(&self) -> bool

Whether this method supports incremental (online) updates.