メニュー
Calculate A/B test statistical significance. Use when: determining if test results are significant; calculating required sample size; estimating test duration; analyzing conversion experiments; making data-driven decisions
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | ab-test-stats |
| description | Calculate A/B test statistical significance. Use when: determining if test results are significant; calculating required sample size; estimating test duration; analyzing conversion experiments; making data-driven decisions |
| license | MIT |
| metadata | {"author":"ClawFu","version":"1.0.0","mcp-server":"@clawfu/mcp-skills"} |
Calculate statistical significance for A/B tests - know when your results are real, not random chance.
| Claude Does | You Decide |
|---|---|
| Structures analysis frameworks | Metric definitions |
| Identifies patterns in data | Business interpretation |
| Creates visualization templates | Dashboard design |
| Suggests optimization areas | Action priorities |
| Calculates statistical measures | Decision thresholds |
pip install scipy numpy click
python scripts/main.py significance --control 1000,50 --variant 1000,65
python scripts/main.py significance --control 5000,250 --variant 5000,300 --confidence 0.99
python scripts/main.py sample-size --baseline 0.05 --mde 0.02
python scripts/main.py sample-size --baseline 0.10 --mde 0.01 --power 0.90
python scripts/main.py duration --traffic 1000 --baseline 0.05 --mde 0.02
# Control: 1000 visitors, 50 conversions (5%)
# Variant: 1000 visitors, 65 conversions (6.5%)
python scripts/main.py significance --control 1000,50 --variant 1000,65
# Output:
# A/B Test Results
# ─────────────────────────
# Control: 5.00% (50/1000)
# Variant: 6.50% (65/1000)
# Lift: +30.0%
#
# Statistical Analysis
# ─────────────────────────
# p-value: 0.089
# Confidence: 91.1%
# Result: NOT SIGNIFICANT (need 95%)
#
# Recommendation: Continue test for more data
# Baseline 5% conversion, want to detect 20% relative lift (1% absolute)
python scripts/main.py sample-size --baseline 0.05 --mde 0.01
# Output:
# Sample Size Calculator
# ──────────────────────────────
# Baseline conversion: 5.0%
# Minimum detectable effect: 1.0% (20% relative)
# Target conversion: 6.0%
#
# Required per variant: 3,842 visitors
# Total required: 7,684 visitors
#
# At 1000 daily visitors: ~8 days
| Term | Definition |
|---|---|
| p-value | Probability result is due to chance |
| Confidence | 1 - p-value (usually want 95%+) |
| Power | Probability of detecting real effect (usually 80%) |
| MDE | Minimum Detectable Effect - smallest lift worth detecting |
| Lift | Relative improvement (variant - control) / control |
| p-value | Confidence | Verdict |
|---|---|---|
| < 0.01 | > 99% | Highly Significant ✓ |
| < 0.05 | > 95% | Significant ✓ |
| < 0.10 | > 90% | Marginally Significant |
| ≥ 0.10 | < 90% | Not Significant ✗ |
category: analytics
subcategory: statistics
dependencies: [scipy, numpy]
difficulty: intermediate
time_saved: 3+ hours/week
Run a structured 5-pass finishing audit on any website before launch — scoring visual polish, technical foundation, UX completeness, content quality, and cross-device readiness on 100 points. Use when: **Pre-launch** - Final validation before going live; **Post-redesign** - Verify nothing broke during the overhaul; **Client handoff** - Structured proof that the site is ready; **Quarterly review** - Catch accumulated debt; **Single-pass focus** - Run just Pass 2 after a perf sprint
Use when rewriting AI-generated text to match a specific person's or brand's authentic voice. Use when AI output sounds generic, corporate, or detectable. Unlike generic "humanizers," this skill requires voice analysis input and produces voice-consistent output, not fake imperfections.
Process large codebases (>100 files) using the Recursive Language Model pattern. Orchestrates parallel sub-agents to map-reduce across files without context rot. Use when: analyzing large repositories; auditing security or auth across many files; finding patterns across 50+ files; processing large log files or data dumps
Model best-case, worst-case, and likely revenue scenarios with sensitivity analysis for strategic planning. Use when: building financial forecasts; presenting board scenarios; planning headcount around revenue uncertainty; modeling pricing changes impact; preparing investor updates with upside/downside ranges
Use when analyzing server logs to understand how AI crawlers (GPTBot, ClaudeBot, PerplexityBot) interact with your site. Use when optimizing content placement for LLM retrieval, diagnosing why AI search isn't citing your content, or auditing crawl patterns to find optimization gaps.
Use when optimizing for AI search visibility where individual experts and entities outperform generic brand pages. Use when building cross-platform authority for people, organizations, or products that AI models should recognize and cite.