Skip to main content

Crate stats_claw

Crate stats_claw 

Source
Expand description

stats-claw — data science on the hot path.

An in-process, zero-runtime-dependency statistical computing library for Rust: probability distributions, hypothesis tests, resampling, optimizers, streaming summaries, and supporting algorithms. The numerics are validated against committed scipy.stats golden fixtures within documented tolerances.

Each distribution is a plain parameter struct (in distributions) over which the behaviour traits (Pdf/Cdf/Quantile/Sample/Moments/ LogCdf) are implemented by hand.

Modules§

algorithms
Unsupervised learning algorithms.
distributions
Probability distributions: parameter structs plus their behaviour traits.
error
Typed error and Result alias for the framework’s fallible numerics.
optimizers
Numerical optimizers minimizing an Objective.
resampling
Resampling-based inference: seeded bootstrap, permutation, and CV splits, plus the percentile confidence-interval estimator built on top of them.
rng
Deterministic pseudo-random number generation and sampling primitives.
special
Special mathematical functions underpinning the distribution numerics.
streaming
Online (streaming) estimators that run in bounded memory.
tests_stat
Statistical hypothesis tests and their effect-size reporting.