StatViz — Visual Statistical Analysis Workbench

StatViz is a browser-based, node-driven data analysis workspace for CSV datasets. It helps users move from raw tabular data to insights, hypotheses, statistical tests, and interpretable result nodes on an interactive canvas.

Capstone Project — Dipan Bag, Spring 2026

---

What It Does

StatViz is designed around a visible analysis graph:

• Dataset node for the uploaded file
• Dataset Summary node for completeness, preview charts, and dataset-level details
• Insight nodes for AI-suggested analytical directions
• Hypothesis nodes for testable claims
• Result nodes for statistical outputs and interpretation
• Next Step / follow-up nodes for continued analysis

Instead of hiding the workflow behind menus, the app keeps the reasoning trail visible.

---

Current Features

• CSV upload in the browser

  • drag and drop
  • click anywhere on the blank canvas to upload
  • use the built-in sample exercise dataset directly from the empty canvas

• Dataset description and summary

  • AI-generated editable dataset description
  • completeness section focused on columns with missing values
  • mixed visual preview cards for numeric and categorical columns
  • More Details branch with dataset-health metrics and short AI focus guidance

• AI-generated insights

  • relationship insights
  • group-difference insights
  • distribution-shape insights
  • outlier-candidate insights

• Hypothesis generation

  • create hypotheses from insight nodes
  • create custom hypotheses manually
  • inline-edit hypothesis statements before testing

• Statistical testing

  • in-browser supported tests via jstat, including:
    • Pearson correlation
    • Welch’s two-sample t-test
    • chi-square test of independence
    • one-way ANOVA
  • AI-assisted fallback when a test is unsupported or estimated

• Result workflow

  • AI-assisted result summaries
  • chart-based result interpretation
  • accept / reject on result nodes
  • re-run test from the result node
  • accepted results can generate a Next Step node and a follow-up editable hypothesis
  • rejected results can generate an alternative sibling hypothesis

• Ask AI

  • dataset-aware right-sidebar assistant
  • can reason over the current graph, results, and branches
  • supports scoped follow-ups through graph context

• Quick analysis summary

  • fixed top-right summary toggle
  • short AI-generated overview of the analysis done so far

---

Tech Stack

Layer	Library / Service
UI framework	React + Vite
Canvas / graph	@xyflow/react (React Flow)
State management	Zustand
Charts	Recharts + custom SVG charts
Statistics	jstat
Layout	@dagrejs/dagre
AI services	OpenAI Chat Completions API
Styling	Plain CSS

---

Running Locally

cd frontend
npm install
npm run dev

Then open:

http://localhost:5173

The active app route is:

/statviz

Note:

• the app uses a user-provided OpenAI API key
• the key is stored in browser session state for the running session

---

Sample Dataset

The empty canvas includes a Use Sample Dataset option.

The app expects the sample exercise CSV at:

frontend/public/sample/exercise/Exercise.csv

This sample is also referenced by the landing page and shared sample-dataset config.

---

Typical Workflow

1. Open StatViz.
2. Upload a CSV or use the sample dataset.
3. Review the dataset description and summary.
4. Open More Details if needed for dataset-health metrics.
5. Generate insight nodes from the summary.
6. Generate or author a hypothesis.
7. Run the suggested test.
8. Review the result node and charts.
9. Accept or reject the result.
10. Continue with a next-step recommendation or an alternative sibling hypothesis.

---

Project Structure

frontend/src/
├── app/
├── pages/
│   └── LandingPage.jsx
├── sampleDatasets.js
├── constants/
├── modes/data/
│   ├── DataModeApp.jsx
│   ├── DataModeApp.css
│   ├── store/
│   │   ├── useDataModeStore.js
│   │   └── analysisContext.js
│   ├── components/
│   │   ├── DataCanvas.jsx
│   │   ├── UploadPopup.jsx
│   │   ├── ApiKeyModal.jsx
│   │   └── ChatPanel.jsx
│   ├── nodes/
│   │   ├── DatasetNode.jsx
│   │   ├── DatasetSummaryNode.jsx
│   │   ├── DatasetDetailsNode.jsx
│   │   ├── InsightNode.jsx
│   │   ├── HypothesisNode.jsx
│   │   ├── CustomHypothesisNode.jsx
│   │   ├── ResultNode.jsx
│   │   ├── NextStepNode.jsx
│   │   ├── InterpretationNode.jsx
│   │   ├── ColumnChart.jsx
│   │   ├── charts/
│   │   │   ├── InsightChart.jsx
│   │   │   ├── ResultChart.jsx
│   │   │   ├── chartData.js
│   │   │   └── charts.css
│   │   ├── nodes.css
│   │   └── index.js
│   ├── api/
│   │   ├── descriptionService.js
│   │   ├── datasetDetailsService.js
│   │   ├── insightService.js
│   │   ├── hypothesisService.js
│   │   ├── customHypothesisService.js
│   │   ├── followupService.js
│   │   ├── analysisSummaryService.js
│   │   ├── chartTypeService.js
│   │   ├── statisticsService.js
│   │   └── chatTools.js
│   └── utils/
│       ├── csvParser.js
│       ├── layoutGraph.js
│       └── mockGraph.js
└── main.jsx

---

Notes

• The app is browser-first: parsing, charting, graph state, and supported statistics happen client-side.
• AI is used for description, insight generation, hypothesis generation, follow-ups, summaries, and interpretation.
• Some result charts and statistical explanation surfaces are still evolving as the visualization system is refined.

---

Deployment

The project is set up for static frontend deployment through GitHub Pages.

The hosted route uses:

/mindmapper/statviz

SPA routing is supported through the 404.html redirect pattern used in the frontend public/ folder.