Claude Statistical Analysis Skill

A Claude Code skill that transforms Claude into a professional statistical consultant. Instead of blindly running whatever method the user requests, it diagnoses first, then analyzes — checking data quality, validating assumptions, and automatically selecting the appropriate statistical method.

What It Does

When you upload a dataset and describe your analysis needs, Claude will:

1. Data Profile (automatic) — Sample size, variable types, missing patterns, distributions, outliers
2. Assumption Checking — Normality, homogeneity of variance, multicollinearity; auto-switches methods when assumptions fail
3. Execute Analysis — Runs the appropriate statistical method with proper controls
4. Output Triple — Every analysis produces three deliverables:
   • APA 7th edition table (Excel + Markdown)
   • Publication-quality figure (300dpi PNG)
   • Results paragraph ready for manuscript (English/Chinese)

Supported Methods

Installation

As a Claude Code Skill

# Copy to your Claude skills directory
cp -r . ~/.claude/skills/statistical-analysis/

Claude Code will automatically detect and activate this skill when you:

• Upload a data file (.xlsx, .csv, .sav)
• Say “help me analyze” / “run statistical analysis”
• Mention a specific method (t-test, regression, SEM, etc.)

R Docker Environment (Optional)

For advanced methods that require R (SEM path diagrams, HLM, IRT):

cd docker/
chmod +x r-stat.sh
./r-stat.sh build    # Build the Docker image (~2GB)
./r-stat.sh test     # Verify installation

Pre-installed R packages: lavaan, lme4, metafor, mirt, psych, tidyverse, semPlot, effectsize, and 30+ more.

Project Structure

.
├── SKILL.md                    # Core skill definition (Claude reads this)
├── references/
│   ├── methods-index.md        # Method selection decision trees
│   ├── code-patterns.md        # Python/R code templates
│   ├── table-formats.md        # APA table format specifications
│   └── full-workflow.md        # Full path detailed workflow
├── docker/
│   ├── Dockerfile              # R environment (rocker/tidyverse + 40 packages)
│   ├── r-stat.sh               # Convenience script for Docker operations
│   ├── README.md               # Docker setup guide
│   └── examples/               # SEM, HLM, meta-analysis R examples
└── assets/
    └── report-template.md      # Analysis report template

Key Design Decisions

“Diagnose Before Analyze”