Super Analyze
Super Analyze is a human-in-the-loop assistant for statistical analysis of experimental datasets.
It turns raw data into a reproducible analysis workflow with automatic detection, explicit review gates, and traceable output artifacts.
English | 中文
Table of contents
What is Super Analyze?
Super Analyze helps research teams move from raw study files to reproducible analysis with better control.
- Detect questionnaire type and design structure from a dataset.
- Keep humans in the loop with required confirmation checkpoints.
- Recommend methods with rationale and alternatives.
- Generate rerunnable scripts and a clean artifact set.
The core value is simple: fewer manual steps, clear decision traces, and lower reproducibility risk.
Feature cards
| Feature | What you get |
|---|
| Smart intake | Detects IPQ, SSQ, SUS, NASA-TLX, or generic experimental datasets, plus subject/condition/DV columns. |
| Two confirmation gates | Mandatory user confirmation on detection and method choice for every analysis path. |
| Method suggestions | Recommends parametric and non-parametric alternatives per dependent variable. |
| Traceable script generation | Produces a readable analyze_<dataset>.py with source labels (rcode vs fallback). |
| Claude-first command flow | Integrated slash-command flow for conversational execution. |
| One command output pack | Exports cleaned data, summary, and figure files together with the script. |
Install
python -m venv myenv
myenv\Scripts\activate
pip install -r requirements.txt
pip install -e .
Register plugin in Claude Code
/plugin marketplace add <YOUR_REPO_PATH>
/plugin install super-analysis@vibe-example-local
Example:
/plugin marketplace add C:/Users/adminroot/Documents/GitHub/vibe_example
/plugin install super-analysis@vibe-example-local
How to use
Recommended (Claude command)
/super-analysis:run text_dataset/ipq.csv
# or shorthand
/super-analysis text_dataset/ipq.csv
Direct CLI usage
python .\scripts\super_analyze.py scan path/to/dataset.csv
python .\scripts\super_analyze.py recommend path/to/dataset.csv
Use the virtual environment interpreter when available:
.\myenv\Scripts\python.exe .\scripts\super_analyze.py scan path/to/dataset.csv
Workflow
Data file → Detect → Confirm → Preprocess → Assumption checks → Confirm → Generate analysis → Export outputs
Phase 1 — Detection (automatic)
- Detect questionnaire type and detected fields.
- Infer design pattern (within/between, single-factor or multi-factor).
- Produce an initial structured report.
Phase 2 — Confirmation 1
- User confirms or corrects the detection report before moving forward.
Phase 3 — Preprocessing
- Apply questionnaire scoring and data cleanup when supported.
- Keep intermediate files deterministic and auditable.
Phase 4 — Assumption checks + recommendation (automatic)
- Compute condition-wise summaries and assumption checks.
- Suggest an analysis method and a fallback for each dependent variable.
Phase 5 — Confirmation 2
- User accepts or replaces the suggested method per dependent variable.
Phase 6 — Script and artifacts
- Generate
analyze_<dataset>.py and result files after confirmations pass.
Method mapping
| Design | Parametric option | Non-parametric option |
|---|
| 2 conditions, within-subject | Paired t-test | Wilcoxon signed-rank |
| 2 conditions, between-subject | Independent t-test | Mann–Whitney U |
| >2 conditions, within-subject | Repeated-measures ANOVA | Friedman |
| >2 conditions, between-subject | One-way ANOVA | Kruskal-Wallis |
| Multi-factor | Two-way ANOVA (or equivalent) | ART or non-parametric alternative |
Outputs
| File | Purpose |
|---|
analyze_<dataset>.py | Traceable analysis script, ready to rerun |
<dataset_stem>_cleaned_scored.csv | Cleaned/scored data table |
<dataset_stem>_analysis_summary.txt | Condensed run summary |
figures/*.png | Auto-generated figures |
Supported questionnaires