## [0.7.2] - 2026-05-03
### Fixed
- `StudentizedRange`: `ArithmeticError` from `:math.pow/2` overflow when `df` is large; outer integrand weight now computed in log-space to avoid intermediate overflow

## [0.7.1] - 2026-05-03
### Fixed
- `StudentizedRange`: `ArithmeticError` raised by `:math.pow/2` overflow when `df` is large (e.g. 301); `c_prefactor/2` now computed entirely in log-space

## [0.7.0] - 2026-03-29
### Fixed
- `TukeyHSD`: confidence interval `level` was always 95.0 regardless of `alpha`; now correctly reflects `(1 - alpha) * 100`
- `StudentizedRange`: passing `inner_nodes: 8` or `outer_nodes: 8` caused a `FunctionClauseError`; unsupported node count guard tightened to 16
- `ANOVA.one_way`: division by zero when all within-group observations are identical (`ms_within = 0`); F-statistic now returns `:infinity` and p-value `0.0`
- `TukeyHSD`: q-statistic and p-value computation guarded against zero standard error
- `TukeyHSD`: missing `effect_size/3` fallback clause for `ms_within = 0` added

## [0.6.1] - 2025-08-23
### Changed
- Modified Tukey's HSD calculation to use much better approximations

## [0.6.0] - 2025-08-23
### Changed
- More compact output format for Tukey's HSD, removed unnecessary fields

## [0.5.1] - 2025-08-23
### Changed
- Replaced the Statistics module with Statistex for mean and median calculation
- Implemented the cumulative distribution function of the F distribution, to replace the one in
  Statistics, since it is not precise enough

## [0.5.0] - 2025-08-21
### Changed
- Better-structured Tukey's HSD result format
- Improved Tukey's HSD effect size calculation

## [0.4.0] - 2025-08-21
### Changed
- Removed alpha from ANOVA parameters, and also removed the 'significant?' result field
- Minor optimizations and refactoring

## [0.3.1] - 2025-08-21
### Changed
- Improved Tukey's HSD p-value estimation

## [0.3.0] - 2025-08-17
### Changed
- Changed Tukey's HSD test result format