Overview

R is a language and environment for statistical computing and graphics, and it is a GNU project. Ross Ihaka and Robert Gentleman started it at the University of Auckland in 1993 as a free implementation of the S language developed at Bell Laboratories, and since 1997 it has been maintained by the R Core Team under the R Foundation. The interpreter, the graphics devices and the package manager all ship together, so the language and the statistics arrive as one install. R is free software under the GNU General Public License (version 2 or later), runs on Linux, Windows and macOS, and the current release is R 4.6.2 “Diversified Portfolio”, published on 29 October 2026.

Key Features

• Statistics in the core language — linear and generalised linear models (lm(), glm()), classical tests (t.test(), chisq.test(), cor.test()), analysis of variance, time-series functions and a model-formula syntax are part of base R, not a library you add.
• Data frames and vectorisation — the table type is a first-class object, and column-wise arithmetic, subsetting with d[rows, cols] and grouping with aggregate() work without a loop; the tidyverse’s dplyr and tidyr add a reworked grammar on top.
• Publication-quality graphics — base graphics and ggplot2 draw plots with mathematical notation, and the device layer writes PDF, SVG, PNG and PostScript without leaving the session.
• CRAN — 25,700 contributed packages at the time of writing, covering machine learning, econometrics, biostatistics, geospatial data and most published statistical methods, installed with install.packages("name").
• Reproducible reporting — R Markdown and Quarto interleave code with prose to produce HTML, PDF and Word documents, and renv locks package versions per project so an analysis still runs a year later.
• Interactive and batch modes — the console for exploration, Rscript file.R for cron jobs and pipelines, and R CMD check for packaging your own code.
• Web apps and APIs — Shiny turns a fitted model into a server-hosted application, and plumber exposes R functions as REST endpoints.
• Interoperability — Rcpp calls C and C++ at native speed, DBI and its drivers talk to PostgreSQL, MySQL and SQLite, reticulate embeds Python, and knitr output lands straight in LaTeX.

Why Use R?

R is what a language looks like when statistics came first. A regression that needs several libraries and a glue layer elsewhere is one function call here, and it returns an object you can print, plot, summarise and test rather than a screen of numbers to interpret by hand. The ecosystem is the other half of the argument: because the people who publish new statistical methods usually publish R code with them, CRAN tends to have the implementation before any other language does, and the packages are peer-reviewed for correctness rather than only for API design. It plays well with an existing stack instead of replacing it — C++ hot loops through Rcpp, Python libraries through reticulate, SQL databases through DBI, LaTeX through knitr — and it is GPL software with identical behaviour on Linux, Windows and macOS. The cost is a few rough edges: base R’s syntax is idiosyncratic, memory-hungry operations are slower than a tuned C or Julia implementation, and a serious project usually needs the tidyverse on day one to stay readable.

Use Cases

• Statistical analysis, hypothesis testing and regression modelling.
• Academic, clinical and epidemiological research, where reference implementations of new methods appear first.
• Cleaning and reshaping tabular data before it reaches a model or a report.
• Publication figures for papers, using ggplot2 or the base graphics device.
• Reproducible analysis pipelines and automatically regenerated reports.
• Time-series forecasting and econometrics, with the forecast, xts and zoo packages.
• Machine-learning prototypes with caret or tidymodels before anything is productionised.
• Shiny dashboards and plumber APIs built directly on an existing analysis.

Getting Started

Install the interpreter from CRAN or from your distribution — sudo apt install r-base on Debian and Ubuntu, the signed .pkg installer on macOS, the Windows installer from CRAN — then check it with R --version. Start the console with R, run a file with Rscript script.R, and quit with q(). Packages install from a CRAN mirror with install.packages("ggplot2"); on Linux, distribution packages for the heavy dependencies (libcurl, libxml2, libssl) save a long compile. RStudio Desktop and the VS Code R extension give the interpreter an editor, but neither is required, and R itself is a separate install from both. The source lives in the R Subversion repository with a read-only GitHub mirror, and the R FAQ documents the licence terms and the supported platforms.

Related Tools

• Python — general-purpose alternative with pandas, NumPy and scikit-learn; stronger for production code, weaker for statistics out of the box.
• Perl — the older text-processing workhorse, still the fastest route through log files and one-liners.