K-Dense skill for hypothesis tests on research data: picks the test, checks assumptions, reports effect sizes with intervals in APA style.
How it works
Foe: Riddle Sphinx, a question your data hasn't answered yet.
Foes
- Riddle SphinxHaze Poison Ward Insight Flicker Pinned Enraged
Party · Context
- AlchemistYour agent100/100Haze Poison Ward Insight Flicker Pinned Enraged
The Riddle Sphinx blocks the way.
What really happensIt stands for a question your data hasn't answered yet.
Alchemist: Context 100/100.
The sphinx asks a riddle. A wrong guess costs dearly.
What really happensAn answer guessed from a hunch instead of a query sends the work the wrong way.
Alchemist takes 11. Alchemist: Context 89/100.
The hero names the riddle before touching a single row.
What really happensThe agent states the hypothesis, outcome, predictors and design first, then looks at n, mean, SD and missing values per group and plots the raw data.
Riddle Sphinx takes 14.
The guide scroll points to the right test for the design.
What really happensIt picks a test from the quick reference or test_selection_guide.md: t-test, ANOVA, chi-square, correlation, regression or a rank method.
Riddle Sphinx takes 17.
The Riddle Sphinx uses Riddle.
Alchemist takes 11. Alchemist: Context 78/100.
Assumptions are tested before the answer is trusted.
What really happensscripts/assumption_checks.py runs normality, variance and outlier screens. The skill says to read them with the design, not switch tests on a p-value.
Riddle Sphinx takes 8. Alchemist: Ward.
Each blow lands with its size and its margin of error.
What really happensEvery test comes with an effect size and a confidence interval, mostly through Pingouin, SciPy and statsmodels.
Riddle Sphinx takes 33.
The result is written up the way reviewers expect.
What really happensResults follow the APA templates: descriptives, exact statistics, effect sizes with intervals and the checks performed.
Riddle Sphinx takes 28. Riddle Sphinx falls.
The Riddle Sphinx falls.
Learn this spell
Claude Code and Codex
gh skill install K-Dense-AI/scientific-agent-skills statistical-analysisNotes
Part of Scientific Agent Skills; gh skill installs this one skill. Targets Pingouin 0.7.0, SciPy 1.18.1 and statsmodels 0.15.0 in a separate uv environment, plus PyMC and ArviZ for the Bayesian examples. The network is needed only to install them, so a sealed trial arena needs those packages already present. Covers t-tests, ANOVA, chi-square, correlation, regression, rank tests and power analysis. For model APIs it points to the statsmodels and pymc skills.
This spell has no recorded sessions in the curator's logs. Tried it? Send a spell report.

