npx claudepluginhub wei-hci/vibe_exampleHuman-in-the-loop statistical analysis plugin on top of rcode. Adds deterministic dataset scan, method recommendation, and traceable script generation workflows.
A Claude Code plugin for statistical analysis of experimental research data. It auto-detects questionnaire types and experimental designs, asks for your confirmation at key decision points, and generates a traceable analysis script.
Built on top of rcode, a statistical helper library in the same repository (see README-rcode.md).
# 1. Install dependencies
python -m venv myenv
myenv\Scripts\activate
pip install -r requirements.txt
pip install -e .
# 2. Register as a local plugin in Claude Code
/plugin marketplace add C:/Users/adminroot/Documents/GitHub/vibe_example
/plugin install super-analysis@vibe-example-local
# 3. Run
/super-analysis:run text_dataset/ipq.csv
# or shorthand
/super-analysis text_dataset/ipq.csv
Data file ──▶ Detect ──▶ Confirm ──▶ Preprocess ──▶ Assumptions ──▶ Confirm ──▶ Analyze ──▶ Output
│ ▲ │ ▲
│ │ │ │
automatic you decide auto-recommend you decide
Phase 1 — Detection (automatic) Scan the file and detect questionnaire type (IPQ / SSQ / SUS / NASA-TLX / generic), independent variables, dependent variables, subject ID, and experimental design (within/between-subjects, single/multi-factor).
Confirmation 1 — You confirm or correct the detection results
Phase 2 — Preprocessing (automatic)
Score questionnaires via rcode functions when available, clean data, and export cleaned_scored.csv.
Phase 3 — Descriptive Statistics & Assumption Checks (automatic) Summarize by condition; run Shapiro-Wilk, Mauchly, Levene, etc. and recommend a statistical method per dependent variable.
Confirmation 2 — You confirm or override the recommended methods
Phase 4 — Main Analysis (automatic)
| Design | Parametric | Non-parametric |
|---|---|---|
| 2 conditions, within | Paired t-test | Wilcoxon |
| 2 conditions, between | Independent t-test | Mann-Whitney U |
| ≥3 conditions, within | Repeated-measures ANOVA | Friedman |
| ≥3 conditions, between | One-way ANOVA | Kruskal-Wallis |
| Multi-factor | Two-way ANOVA | ART |
Phase 5 — Output
| File | Description |
|---|---|
analyze_xxx.py | Traceable analysis script (the core artifact — each block is labeled with its source) |
cleaned_scored.csv | Cleaned and scored data |
summary.txt | Results summary |
figures/*.png | Figures |
Every code block in the generated script is labeled as either rcode-backed or local fallback (using pingouin / scipy / statsmodels).
rcodercode does not yet expose a dedicated wrapper)rcode LibrarySuper Analyze is the orchestration layer; rcode is the statistical library underneath it, handling questionnaire scoring, assumption checks, reporting, and plotting.
For the full list of available functions, see README-rcode.md.