Orientation (1–2 days)
Install & setup: R, RStudio/Posit IDE, Quarto, Git.
Project hygiene: create a project per analysis; use .Rproj, here()/fs, relative paths.
Core mental model: vectors → data frames/tibbles; everything is a vector; recycling rules.
Milestone: You can open RStudio, run code, save a script, knit a minimal Quarto report.
R Fundamentals (1–2 weeks)
Syntax: objects, assignment (<-), atomic types, attributes.
Subsetting: [ ], [[ ]], $ for vectors/lists/data frames.
Vectorization & recycling: write operations over whole vectors; avoid explicit loops initially.
Functions: parameters, return values, lexical scoping; ....
Control flow: if, for, while, repeat, switch.
Base data ops: sum, mean, seq, rep, apply family basics.
Practice: Implement zscore(x), winsorize(x, p), and a group mean function using split, lapply, unsplit.
Milestone: You can write small, reusable functions and index/subset any object confidently.
Data Handling & the Tidyverse (2–3 weeks)
tibbles & readr: robust CSV/TSV import, column types, parsing dates with lubridate.
dplyr core: select, filter, mutate, arrange, summarise, group_by, across.
joins & reshaping: left_join, bind_rows, pivot_longer/wider; factors with forcats.
strings & text: stringr for regex, tokenization basics.
list-columns & rowwise: handling nested data.
Functional patterns: purrr::map_*, error-safe mapping (safely, possibly).
Practice: Clean messy multi-file dataset into a tidy tibble; write a 10-line pipeline with KPIs.
Milestone: You can produce a clean analytic dataset with <20 readable lines.
Visualization Mastery (1–2 weeks)
ggplot2 grammar: aesthetics, geoms, stats, facets, themes.
Scales & guides: continuous vs. discrete, custom palettes, labels.
Communication: annotations, layouts, saving high-quality exports.
Interactivity: intro to plotly, ggiraph.
Practice: Reproduce two published charts and create a multi-panel figure.
Milestone: You can craft publication-ready figures that tell a clear story.
Statistics with R (2–4 weeks)
Descriptives & inference: CIs, t-tests, ANOVA, nonparametrics.
Linear models: lm, diagnostics, interactions.
GLMs: logistic/Poisson with glm.
Resampling: bootstrap, permutation tests.
Multiple testing: p-adjust, FDR.
Effect sizes & reporting: broom (tidy, glance).
Practice: Design a small experiment, analyze with lm/glm, report tidy tables.
Milestone: You can go from hypothesis → model → reproducible report.
Reproducible Research & Reporting (1–2 weeks)
Quarto/R Markdown: parameterized reports, citations, cross-refs.
Project environments: renv.
Pipelines: targets (or drake).
File I/O & databases: readr, arrow, DBI, duckdb.
Versioning: Git basics.
Practice: Convert a one-off script into a targets pipeline with a parameterized report.
Milestone: A “one-click” reproducible project from raw data to final report.
Modeling & Machine Learning (3–5 weeks)
Frameworks: tidymodels suite.
Tasks: regression, classification, CV, hyperparameter tuning.
Feature engineering: recipes for scaling/encoding.
Model selection: metrics, ROC/PR curves, calibration.
Interpretability: variable importance, partial dependence.
Time series: tsibble, fable.
Practice: Build a full ML workflow, evaluate, and report results.
Milestone: You can compare models reproducibly and justify your choice.
Advanced R (3–4 weeks)
Object systems: S3, S4, R6.
Metaprogramming & tidy eval: quosures, {{ }}.
Performance: profiling (profvis), vectorization, data.table.
C/C++ integration: Rcpp.
Parallelism: future, furrr, parallel.
Practice: Rewrite a loop with vectorization or Rcpp and benchmark it.
Milestone: You can diagnose and optimize slow code.
Production, Packaging & Quality (2–3 weeks)
Package development: usethis, devtools, roxygen2, testthat, pkgdown.
Style & linting: styler, lintr.
Shiny apps: reactivity, modules, dashboards.
APIs: plumber; model deployment with pins.
Ops: logging, configs, Docker, CI/CD.
Practice: Publish a small package and deploy a Shiny mini-dashboard.
Milestone: Your work is installable, tested, documented, and shippable.
Domain Tracks (choose 1–2, 2–4 weeks each)
Time series: fable, prophet.
Geospatial: sf, terra, mapview.
Text/NLP: tidytext, topic models.
Bioinformatics: Bioconductor basics.
Finance: quantmod, risk analytics.
Deep learning: torch, keras.
Causal inference: MatchIt, did, fixest.
Milestone: Complete a polished project in your chosen domain.
Debugging, Testing & Teaming (ongoing)
Debugging: browser(), traceback().
Testing: testthat, snapshots.
Data validation: validate, pointblank.
Docs: READMEs, reproducible examples (reprex).
Milestone: You can triage bugs quickly and prevent regressions.
Portfolio Projects (pick 2–3)
Data Story: Clean and visualize public data, publish a Quarto article.
ML Pipeline: Tidymodels + targets pipeline.
Shiny App: Interactive dashboard with real data.
R Package: Helpers + datasets + tests.
Domain Case Study: Time series, geospatial, or text.
Reference Cheatlist
Core: base R, dplyr, tidyr, ggplot2, readr, purrr, lubridate, forcats.
Workflow: here, fs, janitor, renv, targets, arrow, duckdb.
Modeling: tidymodels, xgboost, ranger, catboost.
Viz add-ons: patchwork, ggtext, ggrepel, scales.
Perf: data.table, Rcpp, profvis, bench.
Apps/Prod: shiny, plumber, pins, vetiver.
Suggested 12-Week Plan
Weeks 1–2: Fundamentals + tidyverse.
Weeks 3–4: Visualization + wrangling projects.
Weeks 5–6: Statistics + reproducible research.
Weeks 7–9: Tidymodels + ML project.
Weeks 10–11: Advanced R + optimization.
Week 12: Package or Shiny app.
How to Study
Daily coding practice (45–90 minutes).
Refactor scripts into functions.
Benchmark often.
Narrate your work with Quarto.
Automate with targets.
Share small outputs frequently.
Finishing Line
Ingest messy data, tidy it, model it, and communicate results.
Projects are reproducible (renv, targets) and tested.
Code is idiomatic, optimized when necessary.
Work can be packaged, deployed, and maintained in production. Mastering R is a journey that requires practice, consistency, and building real projects. By following this roadmap, you will progress from learning the basics of syntax and data handling to advanced areas such as modeling, package development, and reproducible workflows. The skills you gain will allow you to analyze complex datasets, create insightful visualizations, and deliver robust, production-ready solutions. Ultimately, this roadmap equips you to not only use R effectively but to think like an R programmer—writing clean, reproducible, and impactful code that adds value in research, business, and beyond.
