Highlights from rstudio::conf 2019: The Premier R Conference
Introduction
The RStudio Conference (rstudio::conf) is the foremost annual gathering for the R community, bringing together data scientists, statisticians, analysts, and developers to explore the latest advancements in R. The 2019 edition, held January 15-18 in Austin, Texas, did not disappoint, featuring a packed lineup of talks, tutorials, and workshops showcasing cutting-edge techniques and tools.
R has cemented itself as a leading platform for data science, with a rapidly growing user base and a vast ecosystem of packages. According to the Tiobe Index, R now ranks as the 8th most popular programming language, up from 18th just four years ago. Much of this growth can be attributed to R‘s expanding capabilities for data manipulation, visualization, machine learning, and reporting.
In this post, I‘ll share my key takeaways from rstudio::conf 2019, diving into the major themes, packages, and insights that I believe will shape the future of R and data science more broadly. Whether you‘re a seasoned practitioner or just getting started with R, there are valuable lessons to be gleaned.
The Exponential Growth of R
A quick glance at the numbers makes it clear that R is experiencing explosive growth:
- The R package ecosystem now encompasses over 14,000 packages, doubling in size every 3 years (source)
- In 2018, the number of R users worldwide surpassed 2 million, with 55% year-over-year growth in the US (source)
- Job postings seeking R skills have increased 500% since 2014 (source)
As Carl Howe noted in his presentation summarizing RStudio‘s survey of R users, this growth is fueled by several converging trends:
- The rise of data science and demand for tools to derive insights from data
- R‘s expanding capabilities and package ecosystem
- Adoption of R in fields like healthcare, finance, tech and government
- Resources making R more approachable (e.g. RStudio Cloud, DataCamp)
"When I first started using R over a decade ago, I never imagined it would reach the audience it has today. It‘s been incredible to witness R‘s evolution from a niche statistical tool to a general-purpose language for data science. The breadth of domains and applications on display at rstudio::conf exemplify this transformation," said Hadley Wickham, Chief Scientist at RStudio.
Key Themes and Takeaways
Powerful New Packages & Frameworks
Much of R‘s success can be attributed to its extensive collection of packages, which simplify and democratize complex data science tasks. Rstudio::conf showcased many notable new additions:
- plumber: Converts R code to web APIs, simplifying productionizing R models and algorithms
- pagedown: Leverages R Markdown to generate publication-quality articles, reports and websites
- tidyverse packages for specialized data wrangling tasks:
These frameworks point to R‘s maturation from a statistical programming language to an all-in-one platform for end-to-end data science workflows. As organizations seek to extract value from data and operationalize machine learning, tools like plumber and pagedown will play a vital role in bringing R to production.
Leveling Up Data Science Skills
With great power comes a learning curve, and several rstudio::conf talks aimed to equip attendees with practical skills for common data challenges:
- Working with categorical data using the forcats package (Amelia McNamara)
- Tidy time series analysis with tsibble and fable (Earo Wang)
- Visualizing uncertainty using Hypothetical Outcome Plots (HOPs) (Claus Wilke)
Uncertainty, in particular, is a critical concept in data science that‘s often glossed over. HOPs offer an intuitive way to represent the distribution of possible outcomes, facilitating communication with non-technical stakeholders. I predict we‘ll see wider adoption of tools like the ungeviz package to visualize uncertainty.
New Techniques for Learning & Teaching R
For those new to R or looking to help others learn, there was valuable advice on effective teaching approaches:
- Teaching data science with puzzles (Irene Steves)
- Empowering a data team with RStudio addins (Hao Zhu)
"Puzzle-based learning encourages you to break down problems into manageable pieces and think algorithmically – core skills for aspiring data scientists," said Irene Steves. Steves‘ "Tidies of March" puzzles not only reinforce key functions from the tidyverse, but promote coding best practices.
Scaling R for Organizations & Teams
With enterprises increasingly leveraging R in production, speakers shared lessons learned on successfully deploying R at scale:
- Using data effectively: StitchFix‘s approach to experimentation, metric design and reproducibility (Hilary Parker)
- Data science as a team sport: Structuring data science teams for impact (Angela Bassa)
- R for better organizational thinking: Applying software engineering best practices to R (Riaz Arbi)
While often overlooked, organizational and process concerns are just as critical as technical ones. Establishing best practices around coding style, testing, documentation and collaboration can make the difference between a thriving, high-impact data science team and an ineffective one.
"Data science teams are most successful when there are clearly defined roles, everyone understands how their work fits into the bigger picture, and there are processes in place to facilitate collaboration and knowledge sharing," said Angela Bassa, Director of Data Science at Chewy.
Deep Learning Comes to R
Despite R‘s extensive machine learning capabilities, it‘s historically lagged behind Python for deep learning. But that‘s starting to change, with a variety of mature frameworks bringing neural networks and deep learning to R:
- Keras: High-level API for building neural networks
- TensorFlow: Open source library for numerical computation and machine learning
- PyTorch: Deep learning framework known for ease-of-use and flexibility
- MXNet: Scalable deep learning library
In her presentation, Sigrid Keydana demonstrated how to implement complex architectures like generative adversarial networks (GANs) and variational autoencoders entirely in R using TensorFlow with eager execution. I expect we‘ll see more R users tackling ambitious deep learning applications in domains like computer vision and natural language processing.
A Bright Future for R
Forecasting the future of any technology is fraught, but if the content at rstudio::conf is a leading indicator, R has a very bright future indeed. The confluence of forces propelling R – a growing user base, maturing ecosystem, and increasing adoption in both academia and industry – show no signs of slowing.
R‘s success is ultimately a testament to its vibrant community. The speakers, attendees, and organizers of rstudio::conf exemplified the curiosity, passion, and spirit of inclusivity that make the R community so special. Their efforts to make data science more accessible and impactful are what will continue to drive R forward.
"The R community is one of the most welcoming and supportive I‘ve ever been a part of. There‘s a real ethos of collaboration and a genuine desire to help more people develop data skills, as evidenced by the many speakers sharing open source teaching materials. I‘m excited to see how the next wave of R users push the boundaries of what‘s possible," said David Robinson, Chief Data Scientist at DataCamp.
Rstudio::conf also highlighted RStudio‘s critical role in stewarding the future of R. From developing essential tools like the RStudio IDE and shiny to investing in education and outreach, RStudio is committed to empowering R users and helping them be successful. I look forward to seeing how offerings like RStudio Cloud and the new RStudio Package Manager continue to accelerate R adoption in the coming years.
Conclusion
2019 is shaping up to be another pivotal year for R and rstudio::conf provided an inspiring look ahead. While the volume and variety of talks made it impossible to see everything, I walked away with an appreciation for the breadth and depth of the R ecosystem, and a conviction that R will play an increasingly vital role in the data science landscape.
But even more than that, I came away energized by the R community and optimistic about what we can achieve working together. The open source ethos, commitment to education, and spirit of inclusivity that animate the R world are a model for what data science can and should be. It‘s not just about extracting insights, but doing so ethically, transparently, and in service of the greater good.
I believe R‘s greatest strength is its community, and if rstudio::conf is any indication, that community is thriving. I‘m grateful to be a part of it, and I can‘t wait to see how the next million R users change the world. Until next year!
