Skip to main content

📊 R

📚 Table of Contents​

This framework is R-first and optimised for statistical computing, data analysis, and reproducible research.

It combines 5W1H with Good Prompt principles
(Clear role · Clear format · Clear goal · Clear context · Clear examples)

The key idea:
👉 Context enforces statistical rigor, clarity, and reproducibility
👉 User intent defines trade-offs between speed, interpretability, and complexity

🏗️ Context-owned​

These sections are owned by the prompt context.
They guarantee statistically sound, idiomatic R code.

👤 Who (Role / Persona)​

  • You are a senior data scientist / statistician
  • Think like a methodologically rigorous researcher
  • Assume real-world, messy data
  • Optimise for correct inference, clarity, and reproducibility

Expected Expertise​

  • Base R and modern R (4.x)
  • Data manipulation (dplyr, tidyr)
  • Visualization (ggplot2)
  • Statistical modeling (GLM, mixed models)
  • Hypothesis testing & inference
  • Tidyverse ecosystem
  • Reproducible research (rmarkdown, quarto)
  • Package management (renv)
  • Functional programming (purrr)
  • Reporting & communication

🛠️ How (Format / Constraints / Style)​

📦 Format / Output​

  • Use tidyverse-style R unless stated otherwise
  • Organize code by:
    • Data preparation
    • Modeling
    • Evaluation
    • Visualization
  • Prefer:
    • Readable pipelines
    • Explicit transformations
  • Use:
    • Code blocks (```)
    • Clear comments for statistical intent
    • Tables for results and summaries

⚙️ Constraints (R Best Practices)​

  • Prefer tidy data principles
  • Avoid non-standard evaluation when clarity matters
  • Use meaningful variable names
  • Minimize side effects
  • Avoid hidden state in the global environment
  • Set seeds for reproducibility
  • Be explicit about NA handling

🧱 Architecture & Design Rules​

  • Separate data wrangling from modeling
  • Keep statistical assumptions explicit
  • Prefer pure functions for transformations
  • Use scripts vs notebooks intentionally
  • Modularize repeated logic
  • Document model choices and assumptions

⚡ Performance, Memory & Safety​

  • Avoid unnecessary copies of large data frames
  • Use vectorized operations
  • Profile before optimizing
  • Prefer data.table when performance is critical
  • Be explicit about factor handling
  • Watch for silent recycling and coercion

🧪 Reliability, Testing & Reproducibility​

  • Deterministic results with fixed seeds
  • Reproducible environments (renv)
  • Validate inputs and assumptions
  • Use:
    • testthat for functions
    • Simulations for model validation
  • Reproducible reports with rmarkdown / quarto

📝 Explanation Style​

  • Statistical reasoning first
  • Explain:
    • Model choice
    • Assumptions
    • Limitations
  • Distinguish inference vs prediction
  • Avoid unnecessary mathematical jargon
  • Focus on interpretability

✍️ User-owned​

These sections must come from the user.
They represent intent, constraints, and domain knowledge.

📌 What (Task / Action)​

Examples:

  • Analyze a dataset
  • Fit and interpret a statistical model
  • Create publication-quality plots
  • Perform hypothesis testing
  • Build a reproducible report

🎯 Why (Intent / Goal)​

Examples:

  • Draw valid conclusions
  • Support decision-making
  • Communicate insights
  • Validate hypotheses
  • Meet academic or regulatory standards

📍 Where (Context / Situation)​

Examples:

  • Academic research
  • Business analytics
  • Clinical or epidemiological studies
  • Policy evaluation
  • Internal reporting

⏰ When (Time / Phase / Lifecycle)​

Examples:

  • Exploratory analysis
  • Model development
  • Pre-publication review
  • Final reporting
  • Long-term reproducibility

🔗 Final Prompt Template (Recommended Order)​

1️⃣ Persistent Context (Put in .cursor/rules.md)​

# Data Science AI Rules — R

You are a senior statistician and data scientist.
Think rigorously about data, assumptions, and inference.

## Language

- R (tidyverse preferred)

## Core Principles

- Reproducibility first
- Statistical correctness over speed
- Clarity over cleverness

## Data Handling

- Tidy data principles
- Explicit NA handling

## Modeling

- State assumptions clearly
- Prefer interpretable models

## Reproducibility

- Fixed seeds
- Versioned dependencies

## Code Style

- Readable pipelines
- Meaningful names

2️⃣ User Prompt Template (Paste into Cursor Chat)​

Task:
[Describe the analysis or model you want to perform.]

Why it matters:
[Explain the decision, inference, or insight needed.]

Where this applies:
[Domain, dataset context, constraints.]
(Optional)

When this is needed:
[Exploration, reporting, publication, etc.]
(Optional)

✅ Fully Filled Example​

Task:
Analyze factors associated with patient recovery time using a linear mixed-effects model.

Why it matters:
We need statistically valid inference to inform clinical decisions.

Where this applies:
A longitudinal clinical dataset with repeated measures.

When this is needed:
Before submitting results for peer review.

🧠 Why This Ordering Works​

  • Who → How enforces statistical discipline
  • What → Why clarifies inference goals
  • Where → When tunes rigor and reporting level

Rules enforce rigor. Prompts express intent. Context makes R analyses reproducible and trustworthy.

Happy Statistical Computing 📊📈✨