Articles about "R"

A recent review by R developers highlights the efficiency of Jarl, a Rust-based linter, in automating code cleanup tasks for R packages. The tool enables detection of unused functions and code smells, supporting more maintainable and reliable software. Automated tools like Jarl are reshaping workflows in software development.

A recent article examines 11 common 'test smells' that can undermine the reliability and clarity of software tests, often leading to undetected bugs or unclear test results. Issues such as hidden dependencies, over-complicated test cases, and reliance on specific test orders are highlighted, alongside practical strategies for improvement. The article provides actionable advice for developers, including the use of dependency injection and clearer test documentation.

CRAN saw the introduction of 376 new packages in April 2026, with the top 40 spanning domains such as machine learning, statistical methods, and computational biology. This curated selection reflects growing innovation in open-source AI tools, statistical computing, and data-driven research, with applications in fields like health, finance, ecology, and programming. Notably, several packages deliver advances in machine learning algorithms and reproducible analytical methods for the R ecosystem.

A new approach leveraging the TabICL tabular foundation model with conformal prediction delivers reliable prediction intervals for tabular data in both Python and R. Experimental results demonstrate comparable performance with Ridge regression, emphasizing practical tools for quantifying uncertainty in machine learning models.

A recent article explores the Atlas-Learn algorithm, a novel approach for learning the structure of data manifolds in high-dimensional spaces. Unlike popular dimensionality-reduction techniques, Atlas-Learn uses concepts from differential geometry to directly reconstruct the intrinsic geometry of data. The implementation and validation are demonstrated in R on the two-dimensional sphere.

Researchers demonstrate the integration of conformal prediction intervals with TabPFN, a pretrained transformer model for tabular data, using both Python and R. The approach provides statistically valid coverage intervals for model predictions without assumptions about the data distribution, yielding robust performance on the diabetes dataset.

A recent tutorial explores the use of the R package 'crossvalidation' for probabilistic time series forecasting, highlighting metrics beyond traditional point forecasts. The package provides tools to assess prediction intervals using empirical coverage rates and Winkler scores, offering a comprehensive validation approach for time series models.

March 2026 saw the inclusion of 371 new packages to CRAN, the comprehensive R software repository. A curated selection of the Top 40 demonstrates ongoing innovation across domains such as causal inference, machine learning, data accessibility, and statistical modeling. These additions reflect R's continued significance in data science, artificial intelligence, and productivity workflows.

A forthcoming webinar hosted by Jumping Rivers will provide a practical guide to using large language models (LLMs) within R and Python environments. The session will address LLM integration, automation, and best practices for data scientists and developers. Key topics include prompt engineering, handling LLM APIs, and responsible AI use.

The new mlS3 R package aims to simplify machine learning in R by providing a unified interface for training and predicting across multiple algorithms. By leveraging R's S3 object system, mlS3 enables users to switch between popular models without changing their code structure. The package targets both classification and regression tasks, supporting major algorithms and the caret package’s vast model selection.