Ruby Statistics Versions Save

What's Changed

• Fix p value in paired T-test by @ylansegal in https://github.com/estebanz01/ruby-statistics/pull/99

New Contributors

• @ylansegal made their first contribution in https://github.com/estebanz01/ruby-statistics/pull/99

Full Changelog: https://github.com/estebanz01/ruby-statistics/compare/3.0.1...3.0.2

Happy coding everyone! 🎉

This change releases a couple of fixes:

• #78 (fixed on #79) which was caused by a liberal usage of round when performing integrals and it was introduced when implemented the workaround hack for BigDecimals.
• #81 fixes an issue that appeared when running builds for Ruby 3.1, which was caused by a bug in BigDecimal. The bug was fixed in version 3.1.2 (which is not included in Ruby 3.1.2), so this gem coerces to rational when calculating the inverse beta function instead of forcing the BigDecimal version as it might cause incompatibilities with previous supported versions.

Additional to that, on #81 we limit again the number of supported ruby versions to be the same as the stable versions published by ruby core: 3.1.2, 3.0.4 and 2.7.6.

Happy coding everyone! 🎉 🎉

This change fixes #34 by coercing BigDecimals to Rational, improving the speed while performing calculations with this class and adding a small improvement in accuracy. 🎉

In theory, this change shouldn't affect in any negative way any dependant code, so feel free to create an issue if something comes up!

This new release includes a small bugfix described in #32 and solved in #33 🎉 !

In this release we have added the quantile function for the Standard Normal as the Inverse Standard Normal distribution! Thanks @dsounded 🎉 !

You just need to instantiate the InverseStandardNormal class and use the cumulative_function method.

Happy stats!

🎉 Happy new year! 🎉

Finally, we closed #17 and now we have available:

Statistical tests

• Spearman's rank coefficient.
• Kolmogorov-smirnov goodness of fit test, also known as two sample ks-test.

Distributions

• Empirical distribution (ECDF).

Soon I'll be updating the documentation with details about how to use it.

Happy statistics to everyone !

:tada: Habemus more discrete distributions! :tada:.

This change address #4 adding the following discrete distributions:

• Negative Binomial Distribution
• Geometric Distribution
• Bernoulli Distribution
• LogSeries Distribution

It also introduces a minor fix to the mean of the beta distribution :grin:. There is a case where alpha and beta can be zero, which is not defined for the calculation.

Happy Stats! Esteban.

This version fixes #23, so we can now have more accurate p-values for two tailored t-tests.

PR #22. Made by @htwroclau.

Fixes some calculation issues with some two sample T-tests and exposes the t_score to the T-test response.

Fixes: #20

This change allow us to know when it's necessary to revisit the specified samples when we have an standard devation of zero, which in practice seems to be an edge case. We are implementing this approach following the answer made to this question where it's important to reconsider samples with std of zero.