rOpenSci is very excited to announce our first peer-reviewed statistical R packages!

One of rOpenSci’s core programs is software peer-review, where we use best practices from software engineering and academic peer-review to improve scientific software. Through this, we aim to make scientific software more robust, usable, and trustworthy, and build a supportive community of practitioners.

Historically, we have focused on R packages that manage the research data life cycle. Now, thanks to work over the past two years supported by the Sloan Foundation we also facilitate peer-review of packages that implement statistical algorithms. The first statistical packages to pass peer review are:

• aorsf: Accelerated Oblique Random Survival Forests , by Byron Jaeger, Nicholas Pajewski, and Sawyer Welden, reviewed by Lukas Burk, Marvin N. Wright, edited by Toby Dylan Hocking
• melt: Multiple Empirical Likelihood Tests by Eunseop Kim, reviewed by Alex Stringer and Pierre Chausse, edited by Paula Moraga
• canaper: Categorical Analysis of Neo- And Paleo-Endemism in R , by Joel H. Nitta, reviewed by Luis Osorio and Klaus Schliep, edited by Toby Dylan Hocking

These packages were peer-reviewed by statisticians and developers to conform to a set of standards we’ve developed with community input. These standards cover areas such as documentation, testing, algorithm design and interoperability. As part of the review process, authors have also annotated their source code to document how and where they comply with these standards.

Evaluating and reviewing statistical packages is a complex and difficult task for editors and reviewers. To help them we’ve also built automated infrastructure for checking submissions for compliance and producing metrics and diagnostics that help both authors and reviewers navigate the review.

We extend an enormous thank-you to the authors, editors, and reviewers who volunteered their time and helped pilot this new system! Also to our statistical advisory board and all the community members who have provided input on our standards and processes.

We’re currently accepting packages in Bayesian statistics, exploratory data analysis, machine-learning, regression, supervised learning, spatial and time-series statistics, dimensionality reduction, clustering, and unsupervised learning. We will be expanding our standards to cover other areas such as network analysis in the the near future.

If you are interested in participating in statistical software peer review at rOpenSci here are some ways to get involved:

• Read about our process and standards (and suggest changes!) in the rOpenSci Statistical Software Development Guide!
• Watch or listen to some of our recent community Calls
• Submit your own package for review or make a pre-submission inquiry
• Volunteer to be a peer-reviewer
• During the next call for applications, apply to our Community Champions program to get support and training in package development and review.