ahsanmandhar-ui/research-data-visualization-designer

Research Data Visualization Designer Skill

Transform raw tabular research datasets (.csv, .xlsx) and statistical model outputs (R broom/tidy, Python statsmodels/scipy, SPSS ANOVA/regression exports) into reproducible, publication-grade statistical figures and analytical dashboards.

All paths below are relative to research-data-visualization-designer/.

---

Core Statistical & Reproducibility Rules

1. Deterministic Computation, Zero Hallucinated Numbers: Compute sample sizes, missingness, group summaries, regressions, and correlation matrices directly from the input .csv / .xlsx or supplied model table. Never invent p-values or drop outliers silently.
2. Mandatory Code Reproducibility: Every run emits a standalone, dependency-minimal reproduce_figure.py script plus reproducibility_manifest.json recording the input dataset's SHA-256 digest and exact software versions.
3. Perceptually Uniform & Colorblind-Safe Encodings: Use Okabe-Ito categorical palettes and cividis / diverging blue-vermilion colormaps (never jet or rainbow).

---

Workflow

1. Profile Dataset: Inspect columns, types, missingness, and summary distributions:

   python scripts/profile_dataset.py --data <path/to/dataset.csv_or_xlsx> --out <work_dir>/dataset_profile.json
   

2. Generate Statistical Figure / Dashboard + Reproduction Script:

   python scripts/generate_dataviz.py \
     --spec <work_dir>/dataviz_spec.json \
     --output-dir <work_dir>/output
   

3. Validate & Verify Standalone Script Execution:

   python scripts/validate_dataviz.py \
     --output-dir <work_dir>/output \
     --spec <work_dir>/dataviz_spec.json
   

4. Visually inspect <work_dir>/output/dataviz_figure.png to confirm distribution clarity, axis labels, and correlation matrix legibility.