stat_power

Native Ruby statistical power analysis and sample-size determination.

Alpha release: stat_power is under active development. The public API may change before the first stable release. Pin an exact version in production or research environments where reproducibility matters.

The current release is a prerelease:

gem install stat_power --prerelease

or pin the exact alpha:

gem install stat_power -v 0.1.0.alpha.3

With Bundler:

gem "stat_power", "0.1.0.alpha.3"

The first compatibility target is the established CRAN pwr package. The implementation is native Ruby and is validated against published formulas and reference numerical results rather than being a line-by-line source port.

Implemented compatibility currently includes:

• pwr.norm.test
• pwr.p.test
• pwr.2p.test
• pwr.2p2n.test
• pwr.t.test
• pwr.t2n.test
• pwr.r.test
• pwr.anova.test
• pwr.f2.test
• pwr.chisq.test
• ES.h
• ES.w1
• ES.w2
• power-curve data generation equivalent in purpose to plot.power.htest

The library also contains the numerical distribution machinery required by these methods, including central/noncentral t, F, and chi-square distributions.

See docs/pwr_parity.md for the full compatibility matrix.

require "stat_power"

result = StatPower::TTest.two_sample(
  effect_size: 0.5,
  alpha: 0.05,
  power: 0.8
)

result.sample_size
# continuous observations required per group

result.required_sample_size
# smallest whole-number sample size per group

Power-analysis methods follow a common convention: one principal parameter is omitted and solved from the remaining values.

For example, achieved power:

result = StatPower::Correlation.solve(
  correlation: 0.3,
  sample_size: 50,
  alpha: 0.05
)

result.power

• parity with the statistical families provided by CRAN pwr
• idiomatic Ruby APIs
• sample-size determination and achieved-power calculations
• inverse power problems
• effect-size utilities
• explicit assumptions and numerical tolerances
• independently reproducible numerical validation
• later extensions beyond pwr

During the 0.1.0.alpha.* series:

• numerical correctness and validation take priority over API stability
• method and result names may change when inconsistencies are found
• every implemented statistical family should include independent reference tests
• new CRAN pwr parity targets may be added between alpha releases

The transition to beta will indicate that the public API is approaching a freeze candidate.

Set up the checkout:

bin/setup

Open a console with the library loaded:

bin/console

Run everything CI runs:

bundle exec rake verify

That is the same as running each check individually:

bundle exec rspec          # specs
bundle exec rubocop        # lint
bundle exec rbs validate   # signature syntax
bundle exec steep check    # lib/ type checked against sig/

Measure test coverage. Instrumentation roughly doubles the runtime, so it is opt-in; CI enforces a minimum in a dedicated job.

COVERAGE=1 bundle exec rspec
open coverage/index.html

Build the gem locally:

gem build stat_power.gemspec

Contributions are welcome. Because this is a numerical library, the bar is reproducibility: every statistical result must be justified by a published formula or an independently generated reference value.

See CONTRIBUTING.md for the workflow, the shape a new power-analysis family is expected to take, and how to supply reference values.

Participation is governed by the Code of Conduct.

If stat_power disagrees with pwr, R, G*Power or a textbook, that is the highest-priority class of bug here. Open an issue using the Numerical discrepancy template.

See SECURITY.md for supported versions and how to report a vulnerability privately. A numerically incorrect result is a bug, not a vulnerability; report those as ordinary issues.

See docs/mathematical_conventions.md.

For numerical reference fixtures and tolerance rules, see docs/validation.md.

See ROADMAP.md.

See RELEASING.md.

MIT.