Claude Skills
What’s a Claude skill?
A skill is a small folder of instructions and reference material that Claude reads before helping you with a particular kind of task. You don’t need to know anything about AI tooling to use one: install it once, then just describe your data and your question. Claude notices when the skill applies, loads it, and follows the book’s process.
The skills are included with the book. Buying the book — or signing in, if you already have — unlocks the download links below, along with Sections 3–6, the appendices, and Statifier. They’re a study aid, not a substitute for understanding why an analysis works — the book sections behind each skill are where that understanding comes from.
Skill 1: Choosing the Right Statistical Test
Download the skill (zip) Included with full access — you’ll be asked to sign in.
A companion to Section 5.1. Claude walks the book’s 20-question decision tree with you — usually just 2–4 conversational questions, since it can infer most branches from your description — and recommends a test, tracing exactly which branch got you there. Then it checks the test’s assumptions (independence, normality, equal variances) with you, suggests a nonparametric alternative if one fails, and offers runnable R code using your own variable names.
Just describe your situation the way you would to a stats-savvy friend:
“I surveyed 40 students before and after a workshop and measured their confidence on a 1–100 scale. Did the workshop change anything?”
Claude will recognize this as a test-selection question, follow the tree to a paired t-test, explain why, and walk you through the assumptions. When the right answer is a test the book doesn’t cover (like McNemar’s or Fisher’s exact test), Claude names it anyway and tells you it’s outside the book.
What’s inside:
choosing-the-right-statistical-test/
├── SKILL.md (how Claude should interview you)
├── READER-INSTRUCTIONS.md (this page, in brief)
└── references/
├── decision-tree.md (the book's 20-question decision tree)
└── assumptions.md (assumption checks + nonparametric alternatives)Skill 2: Conduct a Statistical Hypothesis Test
Download the skill (zip) Included with full access — you’ll be asked to sign in.
A companion to Chapters 3.3–3.5. It picks up where test selection leaves off: you know (or think you know) which test to run, and now you want to run it properly and report it like a researcher, not a spreadsheet. Claude walks Dr. R’s 7 Steps in order — checking assumptions first, helping you phrase a proper research question and pick one alternative hypothesis, justifying your choice of α in terms of cost, risk, and ethics (not just “everyone uses 0.05”), calculating the test statistic analytically, finding the p-value, drawing the conclusion, and confirming everything in R with a confidence interval.
At the end, Claude verifies that the analytical answer, the R answer, and the confidence-interval conclusion all agree, then writes up your results as a succinct research report: Abstract, Introduction, Background, Methodology, Results, Discussion and Conclusions, References. If you’re still planning your study, it can also calculate the sample size you need with a power analysis (the pwr package, aiming for power ≥ 0.80).
For example:
“28 employees rated job satisfaction before and after our training program, on a 1–10 scale. Walk me through a paired t-test and write it up.”
What’s inside:
conduct-a-statistical-hypothesis-test/
├── SKILL.md (how Claude should guide you)
├── READER-INSTRUCTIONS.md (this page, in brief)
└── references/
├── seven-steps.md (Dr. R's 7 Steps + the report structure)
├── hypothesis-templates.md (RQ templates and H0/Ha tables per test)
└── power-analysis.md (sample sizes with pwr; Type I/II errors)Better together
Install both skills and Claude hands off from one to the other: first it picks the test with the Section 5.1 decision tree, then it runs the test with Dr. R’s 7 Steps and writes the report. That’s the same arc the book teaches — and the same one the 7-Step Wizard and Statifier follow here on the site.
Installing a skill
The steps are identical for both skills — download the zip above, then:
- In Claude, open Settings → Capabilities and find the Skills section. (You need a plan that supports skills; if you don’t see the section, check that “Code execution / Skills” is enabled.)
- Click Upload skill and select the zip file — no need to unzip it.
- That’s it. Claude will automatically use the skill whenever your question is about choosing or running a statistical test.
- Unzip the folder.
- Move it into your personal skills directory so it’s available in every project:
- macOS/Linux:
~/.claude/skills/choosing-the-right-statistical-test/ - Windows:
C:\Users\<you>\.claude\skills\choosing-the-right-statistical-test\
conduct-a-statistical-hypothesis-test) - macOS/Linux:
- Restart Claude Code. Trigger a skill explicitly by typing
/choosing-the-right-statistical-testor/conduct-a-statistical-hypothesis-test, or just ask your question naturally.
Upload the skill with the Skills API and attach it to your requests via the code-execution container. See Anthropic’s documentation at https://docs.claude.com/en/docs/agents-and-tools/agent-skills for current details.
A note on trust
The skills constrain Claude to the logic printed in this book — the same questions, the same branches, the same rules (you can reject or fail to reject a null hypothesis; you can never accept one). Where the decision tree honestly says “I’m not sure what you’re trying to do,” Claude says so too, rather than inventing a recommendation.
Two caveats worth repeating from the book itself: this is not a one-size-fits-all procedure — your data can often legitimately be analyzed in more than one way — and hypothesis testing is a game of disproving, not proving. Read the corresponding book sections before you report results, and you’ll know not just what Claude recommended, but why it’s right.