Like ministat, but smaller?
A Rust library and CLI tool for evaluating whether two or more sets of measurements are statistically different. It does
this by performing a Welch's t-test at a particular confidence level, making it suitable for small sets of
measurements (e.g., multiple runs of a benchmark). It's inspired largely by FreeBSD's
ministat (written by
$ nanostat examples/iguana examples/leopard examples/chameleon examples/leopard: Difference at 95% confidence! 643.50 > 300.00 ± 293.97, p = 0.026080480978720635 examples/chameleon: No difference at 95% confidence.
cargo install nanostat --features cli
Copyright © 2021 Coda Hale
Distributed under the Apache License 2.0 or the MIT license.