borghei/Claude-Skills / data-analytics/statistical-analyst
statistical-analyst
VerifiedApplied statistics for business and product questions — test selection, assumption checks, power planning, effect sizes with intervals, multiplicity correction. Use when interpreting an experiment, sizing a study, or vetting a claim.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 15.6 KB |
| assets/experiment_design_template.md | markdown | 3.3 KB |
| assets/sample_contingency.json | config | 424 B |
| assets/sample_experiment.json | config | 240 B |
| assets/sample_question.json | config | 468 B |
| assets/sample_revenue.json | config | 1.3 KB |
| assets/sample_session_times.json | config | 938 B |
| references/effect-sizes-and-communication.md | markdown | 8.7 KB |
| references/test-selection-and-assumptions.md | markdown | 10.1 KB |
| scripts/run_test.py | script | 5.8 KB |
| scripts/stats_core.py | script | 8.7 KB |
| scripts/test_impl.py | script | 12.7 KB |
| scripts/test_selector.py | script | 14.6 KB |
assets/experiment_design_template.md
View on GitHub# Statistical Design Plan — <study name>
> Fill this in **before** collecting or looking at data. A plan written after
> the fact is not a plan, and every section below exists because filling it in
> late is where the study goes wrong.
**Owner:** <name>
**Date written:** YYYY-MM-DD
**Decision this informs:** <the specific action that will be taken>
---
## 1. Question
**Business question:** <plain language, one sentence>
**Statistical question:** <restated as a comparison, association, or trend>
**What we will do differently depending on the answer:** <if the answer changes
nothing, stop here and do not run the study>
---
## 2. Primary metric
| Field | Value |
|-------|-------|
| Metric name | |
| Definition (exact) | |
| Unit of measurement (user / session / order) | |
| Outcome type (binary / continuous / count / ordinal) | |
| Baseline value | |
**One primary metric only.** Secondary metrics are listed in section 6 and are
explicitly exploratory.
---
## 3. Decision threshold
| Field | Value |
|-------|-------|
| Minimum effect worth shipping (MDE) | |
| How that was derived | <cost to ship vs value per unit> |
| Direction of interest | one-sided intent / two-sided test |
Set before the data arrive. This is the number that turns the result into a
decision instead of an argument.
---
## 4. Design and sizing
| Field | Value |
|-------|-------|
| Randomized? | yes / no — if no, name the confounders |
| Assignment unit | |
| Groups and allocation | |
| Alpha | 0.05 |
| Power | 0.80 / 0.90 |
| Required n per group | <from test_selector.py> |
| Expected time to reach n | |
| Fixed horizon or sequential? | |
If the achievable n is below the required n, say so **here**, before running.
An underpowered study should be a conscious decision, not a discovery.
---
## 5. Test and assumptions
| Field | Value |
|-------|-------|
| Planned test | <from test_selector.py> |
| Why this test | |
| Fallback if assumptions fail | |
| Assumption checks to run | independence / skew / outliers / variance ratio |
**Independence check (mandatory):** how will you confirm one row per unit?
Clustering is the most common and most damaging violation, and almost nobody
checks it.
---
## 6. Secondary and guardrail metrics
| Metric | Type | Role | Correction |
|--------|------|------|------------|
| | | exploratory / guardrail | Benjamini-Hochberg |
Number of comparisons in the family: **<m>** → corrected alpha: **<alpha/m>**
---
## 7. Stopping rule
- [ ] Fixed horizon: we will look once, at n = <N> or date <D>
- [ ] Group sequential: interim looks at <points> with <spending function>
- [ ] Always-valid: confidence sequences, continuous monitoring
**We will not** monitor a fixed-horizon test and stop on significance.
---
## 8. Pre-specified subgroups
| Subgroup | Why specified in advance |
|----------|--------------------------|
| | |
Any subgroup not listed here is exploratory and will be described as such.
---
## 9. Reporting commitments
- [ ] Effect stated in domain units with a confidence interval
- [ ] Outcome shared regardless of direction or significance
- [ ] Deviations from this plan disclosed in the body, not a footnote
- [ ] If null: the upper bound of the effect we could have missed is stated
---
## Deviations log
| Date | What changed | Why | Approved by |
|------|--------------|-----|-------------|
| | | | |
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |