Struct RunningLogMoments 

pub struct RunningLogMoments { /* private fields */ }

Running mean/variance of ln(x) for positive quantities such as step sizes.

This is a numerically-stable Welford moment tracker on the log scale — not a Bayesian posterior (the conjugate model for a log-normal with unknown mean and variance would be Normal-Inverse-Gamma). It is named to say exactly that.

Unlike the previous LogNormalPosterior, the variance is a clean population variance with no prior contamination: after a single observation the variance is 0, and there is no spurious +1.0 sum-of-squares term injected at n = 2.

Implementations

impl RunningLogMoments

pub fn new() -> Self

Create an empty tracker (no observations, no prior).

pub fn observe(&mut self, x: f64)

Observe a positive value (non-positive values are ignored).

pub fn count(&self) -> usize

Number of observations.

pub fn mean_log(&self) -> f64

Mean of ln(x).

pub fn var_log(&self) -> f64

Population variance of ln(x) (M2 / n, and 0 when n < 1).

pub fn sample_var_log(&self) -> Option<f64>

Unbiased sample variance of ln(x) (M2 / (n − 1), None when n < 2).

pub fn mean(&self) -> f64

Mean in the original space, exp(μ + σ²/2).

pub fn mode(&self) -> f64

Mode in the original space, exp(μ − σ²).

pub fn sample<R: Rng>(&self, rng: &mut R) -> f64

Draw a log-normal sample using the current moment estimates.

Trait Implementations

impl Clone for RunningLogMoments

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for RunningLogMoments

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for RunningLogMoments

fn default() -> Self

Returns the “default value” for a type.