Keynotes and Talks at posit::conf(2023)

R users don't always come in teams. Often, you may be the only useR on the block. But, one miraculous day, you get to welcome more R folks on your team. Suddenly, the little R system you created to suit your needs, like a custom R package, styling, and file organization, isn't just for you. Want to suddenly overhaul that one package you wrote two years ago? It probably won't work when your colleagues try to update it. Your new teammates are data.table fans, but you prefer the tidyverse. Do you need to refactor? Are style choices, like indentation important when collaborating, or are you just being persnickety? In this talk, you will learn how to bring new teammates onboard and blend your respective styles without pulling your hair out.

I’d like to a tell the good people at conf about how we deal with the fact that my boss has too many good ideas for data science projects. Specifically, we use an adapted version of the Eisenhower Matrix that lays out our projects according to the effort required and value they will produce. Given the functionally unlimited number of data science projects that a team could do, I think the audience could benefit from this example of how we keep our team focused on valuable work while also reducing the stress that comes from never ending list of projects.

How does a risk adverse Pharma Biostats organization with 600+ people switch from using proprietary software to using R and other open-source tools for delivering clinical trial submissions? First slowly then all at once. GSK started the transition of using R for its clinical trial data analysis in 2020 and now uses R for our regulatory reviewed outputs. The AccelerateR Team, an agile pod of R experts and data scientists, rotates through GSK Biostats study teams sitting side by side to answer questions and mentor during this transition. We will share our experience from AccelerateR and how other organizations can use our learnings to scale R from pilots to full enterprise adoption and contributing to open-source industry R packages.

In March 2022, I was tasked by the Royal Statistical Society with creating a new online publication: a data science website for data science professionals. I've been a print journalist for 20 years and have worked on websites in that time, but my coding ability began and ended with wrapping "<a href=" tags around text and images. That is, until I discovered Quarto. In this talk, I'll describe how I explored, learned and fell in love with the Quarto publishing system, and how I used it to build a website -- Real World Data Science -- from the ground-up. I'll also make the case for why journalists and others with an interest in science communication should discover what Quarto can do -- and what they can do with it!

Traditional Quarto documents often lack interactivity, limiting the ability of students and researchers to fully explore and engage with the presented topic. In this paper, we propose a novel approach that utilizes webR, a WebAssembly-powered version of R, to seamlessly embed R code directly within the browser without the need for a server. We demonstrate how this approach can transform static Quarto documents into dynamic examples by leveraging webR's capabilities through standard quarto code cells, enabling real-time execution of R code and dynamic display of results. Our approach empowers educators and researchers alike to harness the power of interactivity and reproducibility for enhanced learning and research experiences.

You’ve deployed an API to production, but how do you know when non-zero exit codes occur? Our team has 10 APIs in production used in our organization. They are hosted in Posit Connect, but we also have APIs running in Docker containers deployed on Azure. As of today, Posit Connect is limited in its ability to notify when an API encounters an unexpected error. We’ve introduced into our Plumber-based APIs {log4r} and append our logs to syslog, which are routinely captured by Azure Monitor and made available in the cloud. By setting up alerts, we now get notified by email -- and text messages for urgent issues – for non-zero exit codes with descriptive error messages. With this solution, we have more visibility into the health of our APIs

Data scientists are creating incredibly useful data products at an accelerating rate. These products are consumed by others who expect them to be accurate reliable and timely, often promises unfulfilled. In this talk we will explore how to use common CI/CD pipeline tools already within reach of attendees to automatically test and deploy their apps, APIs and reports.

Are you considering or curious about developing code-based tools for scientists? Whether you are an experienced developer or a fellow Posit Academy graduate who might be stepping into this role for the first time, the aim of my story is to inspire you and help you navigate this process. While developing custom R functions, packages, and Shiny apps for diverse analytical capabilities and users in R&D, I learned why it’s important to collect certain information at the start before writing any tidying, analysis, visualization, and web application code. In this talk, I will share the essential technical questions that help me define and plan for success.

A 5-minute talk to discuss how I’ve used Quarto and Bootstrap variables to quickly make Shiny’s new website look as it should. The Quarto user I have in mind works at an organization with specific brand guidelines to follow. I’ll discuss how to set up your theme, show some key Quarto settings, and how Bootstrap’s Sass variables are your best friend.

Slow dashboards lead to a poor user experience and cause users to lose interest, or even become frustrated. A common culprit of this situation is a slowly rendering plot. During the talk, we will dive deeper into how plots are rendered in Shiny, identify common bottlenecks that can occur during the rendering process, and learn various techniques for improving the speed of plots in R/Shiny dashboards. These techniques will range from more efficient data processing to library-specific optimisations at the browser level.

Using Small-Multiples (faceted graphs) is an effective way to compare patterns across many dimensions. In this talk, I'll walk you through some ways to layout your individual facets according to the underlying data. For example, maybe each facet represents a city or point on a 2D plane - we'll explore ways to organize facets in a grid that mimics the data itself - unlocking your ability to explore patterns in 4+ dimensions. Other solutions to this problem rely on manually-curated lists that map common layouts to a grid, but in this talk we'll explore solutions that work on EVERYTHING. I'll show you how to incorporate this technique into your viz, & how I built the libraries, since there are some interesting Data Science concepts at play.

As data scientists, we care about making valid statistical inferences from experiments. And we've adapted well-established and well-understood statistical methods to help us do so in our A/B tests. Our stakeholders, though, care about making good product decisions efficiently. I'll describe how the way we design A/B tests can put these goals in tension and why that often causes misalignment between how A/B tests are intended to be used and how they are actually used. I'll also talk about how I've used R to implement alternative experimental approaches—including non-inferiority designs and value-of-information analyses—that have helped bridge the gap between data scientists and stakeholders.

Maintaining machine learning models in production can be quite different from maintaining general software engineering projects, each with different challenges and common failure modes. In this talk, learn about model drift, the different ways the word “performance” is used with models, what you can monitor about a model, how feedback loops impact models, and how you can use vetiver to set yourself up for success with model maintenance. This talk will help practitioners who are already deploying models, but this is also useful knowledge for practitioners earlier in their MLOps journey because decisions made along the way can make the difference between resilient models that are easier to maintain and disappointing or misleading models.

Conformal inference theory enables any model to produce probabilistic predictions, such as prediction intervals. We'll demonstrate how these analytical methods can be used with tidymodels. Simulations will show that the results have good coverage (i.e., a 90% interval should include the real point 90% of the time).

Say goodbye to installation headaches and hello to a universe of possibilities with R-Universe! Take your R package development to new frontiers by organising and sharing packages beyond the bounds of CRAN. R-Universe's reliable package building process strengthens installation and usage instructions, resulting in fewer support requests and an easy installation experience for users. With webpages and an API for exploring packages, R-Universe creates a streamlined and tidy ecosystem for R-package constellations. Also, you can build a custom toolchain for your users, relieving your workload and empowering users to help themselves. Join me to learn how to explore the vastness of R-Universe and expand your package development possibilities!

With the introduction of Shiny for Python in 2022, users can now use the power of reactivity with their favorite python packages. Shiny can be used to build interactive reports, dashboards, and web apps, that make sharing insights and results both simple, and dynamic. This includes apps to display and explore popular Machine Learning models built with staple python packages like pandas, scikit-learn, and TensorFlow. This talk will demonstrate how to build simple Shiny for Python apps that interface with these packages, and discuss some of the benefits of using Shiny for Python to build your web apps.

We’re often helping developers to assess, fix and improve their shiny apps, and often the first thing we do is see if we can deploy the app. If you can’t deploy your shiny app, it's a waste of time. If you can deploy it successfully, then at the very least it runs, so we’ve got something to work with. There’s a bunch of reasons why apps fail to deploy. They can be easy to fix, like Hardcoded secrets, fonts or missing libraries. Or they can be intractable and super frustrating to deal with, like manifest mismatches, resource starvation and missing libraries. At the end of this talk I want you to know how to identify, investigate and proactively prevent shiny app deployment failures from happening. See you there!

duckdb is an analytical database system that works great with R and Python and other host systems. dplyr is the grammar of data manipulation in the tidyverse, tightly integrated with R, but it works best for small or medium-sized data. The former has been designed with large or big data in mind, but currently you need to formulate your queries in SQL. The new duckplyr package offers the best of both worlds. It transforms a dplyr pipe into a query object that duckdb can execute, using an optimized query plan. It is better than dbplyr because the interface is "data frames in, data frames out", and no SQL code is generated. The talk presents first results, a bit of the mechanics, and an outlook for this ambitious project.

How do you create brand cohesion across your large team when it comes to data viz? My colleague and I, inspired by the BBC’s bbplot, developed a package on top of ggplot2 to create a common look and feel for our team's products. This effort improved not just the cohesiveness of our work, but also trustworthiness. By creating this package, we reduced the reliance of using defaults and the time spent each project customizing numerous graphic elements. More importantly, this package provided an easier on ramp for new teammates to adopt R. We would like to share our journey within a federal agency developing a style guide, to guide and inspire other organizations who could benefit from developing their own branding package and guidance.

Quarto is a powerful engine to generate documents, slides, book, websites and more. The default aesthetics looks good, but there are times where you want and need to change how they look. This is that talk. Whether you want your slides to stand out from the crowd, or you need your ducuments to fit within your corporate style guide, being able to style quarto documents are a valuable skill. Once you have persevered and created the perfect document, you don't want the effort to go to waste. This is where templating comes in. Quarto makes it super easy to turn a styled document into a template to be used over and over again.

RMarkdown and Quarto are shifting the paradigm for how professionals learn to code, write documentation, and teach others. I'll showcase the style I first experienced by Jenny Bryan (Stat545, Happy Git With R, What They Forgot): narrative and code together, shared openly as a website, that builds trust with learners and provides a resource to consult during and outside of live-teaching. This mode shifts culture in a powerful way, since we can fork this idea, repeat it in our own work. I’ll share examples of how professional researchers are using this style to write open documentation, from technical workflows to community onboarding. I’ll also highlight design elements for writing open documentation, and tips for doing so with Quarto.

In education research, knowledge recall is often considered the easiest aspects of learning, and creation of new knowledge the most challenging. Skill development therefore often requires learners to consistently ‘do’ the skill and receive feedback from experts to allow them to fully enter a new culture of practice. This is particularly challenging for those who are interdisciplinary & have limited control over their workload, such as medics and field scientists. In this talk, an educational scientist describes the previous 10 years of supporting veterinary scientists to adopt Open Science practices surrounding Data Science. What worked, what failed miserably, and reflections on why it can be so hard to get a horse to drink.

Everyday we are losing data scientists, not because people aren’t excited but because they walk into the process and become stuck and they don’t have the confidence or the resources to move further along their journey. At The Carpentries we train novice scientists and researchers how to comfortably move from dipping their toe in the water to successfully going deeper to obtain all of the resources afforded to them. I'd like to tell my story as a Data Scientist and offer resources and tips to build confidence to those who are new to their journey. The tools are available however, it is not always easy to find them. The participants must be willing and ready to keep going! I'm the lifeguard ready to help!

Running Connect with off-host content execution in Kubernetes is very cool and allows you to enable some powerful and sophisticated workflows. The question is, do you really need it? How do you evaluate and decide? Let’s have a candid conversation about whether Connect content execution on Kubernetes is right for you and your organization. Moving to Kubernetes will introduce complexity. What do you hope to get in return? If your goals begin and end with cost reduction and ease of scaling, there are potentially very good alternatives to consider. It would be great if scaling Connect horizontally were easier. We’ll talk about why auto scaling Connect (specifically scaling down) is a tricky problem and what we can do about it.

Through my participation in R4DS Online Learning Community, I have advanced my R and data science skills, making me a better student than I otherwise would have through just my studies. As a non-traditional MS Statistics student with an undergraduate background in economics, I had absolutely no experience with the R programming language prior to pursuing my Master’s degree. In July 2021, with hopes of getting a headstart on learning R before beginning my degree program, I joined the R4DS Slack Workspace. Along with helping to improve my programming skills, R4DS has connected me with scholarship, mentorship, and other opportunities, and I think that it would be beneficial for other students to know about this great resource.

Onboarding new hires can be a challenging process, but taking a problem-focused approach can make it more meaningful and rewarding. In this talk, I will share how I discovered the ultimate new hire hack by creating two small packages that gave me the confidence I needed when I started at BMS. Through building these packages, I not only learned R things like using bslib and making font files available for published dashboards, but also gained a deep understanding of my company's internal systems and workflows, and connected with my team via lots of questions. The resulting packages are still heavily used today. Join me to discover how small packages can have a broad impact and what hiring managers can do to help.

We build a great case on how to combine R and Python in a production environment. So the justice department's back office monitors the smooth processing of all traffic fines in Belgium. They gather that data from all police departments. On this data, they check if any anomalies occur. The back-office monitors that using a shiny application where they can see traffic signs showing the status of the whole operation and the status is built using Python scripts that perform anomaly detection if the number of fines is inline what they expect daily. And the results of those checks are delivered to a front-end shiny application with Python flask API.

The development of software can be costly and time-consuming. If the end users are not involved in the process from the start the tool we built may not meet their needs. In this presentation, I will discuss how prototyping in R Shiny can help you build the right tool and save you from spending millions of dollars on a tool no one will use. I will explore the advantages of using R Shiny for prototyping, particularly its ability to rapidly build interactive applications. I will also discuss how to design effective prototypes, including techniques for gathering user feedback and using that feedback to refine your tool. I will emphasize the importance of presenting real-life data, particularly when building data-driven tools.

There are many resources that can help you design better apps. But what if your org creates many apps? Scaling good design to larger groups dials the challenge up to 11. In this talk I will share how we approach the problem at Appsilon. I will present our in-house Design Guide. I will share the successes and failures we've had building it and helping a wide variety of developers use it. I will then share some tips about what you might want to consider if you want to help your org build better apps at scale.

A big part of being in the R community is sharing your knowledge in different forums, no matter how big or small. So what better way to do that than a blog? And what better way than using R and RStudio to build and maintain that blog and website? This was the route I took to challenge myself in putting myself out there more in the community to find my voice, share my knowledge and learn new things. In this talk, I will reflect on lessons learned and gains I have spent the past year blogging and sharing my website for all to see. The side effects include professional and personal benefits – from a clear profile of my skills to the progression of the development of my art. You may leave inspired to try the challenge for yourself.

The transition from analyzing data in R to making packages in R can feel like a big step. Writing code to clean data or make visualizations seems categorically different from building robust “software” on which other people rely. In this talk, I’ll show why that distinction is not necessarily true by discussing my personal experience from learning R in graduate school to reporting bugs on GitHub to becoming a co-author of the tidycensus package and a practicing data scientist. The positive and supportive R community on GitHub, Twitter, and elsewhere contribute to why anyone who writes R code can become a package author. * I have not actually gotten rich but I did get freelance data work based on my package contributions!

I am a biologist wearing R package developer shoes. I recall writing my first package was an intimidating task. Here, I will share the parallels between package dev and biology that helped me through process. Like a food web, every file has a unique role and is interconnected with other components. Experimenting with one will naturally result in change in another. Like the internal physiology of our bodies, routine testing ensures the package is in perfect equilibrium with its environment. Much like the forces that promote species adaptation, addressing bugs improves the functionality of the package. I hope sharing my perspective will help others see package development as wonderful as the natural world and dispel any hesitation to start!

A few years ago my company made a sudden shift from R to Python which was quite bad for my career because I didn't really know Python. The main issue was that I couldn't find a niche which allowed me to use my existing knowledge while learning the new language. Shiny for Python is a great niche for R users because none of the Python web frameworks can do what Shiny can do. Additionally almost all of your knowledge of the R package is applicable to the Python one. This talk will provide an overview of the Python web application landscape and articulate what Shiny adds to this landscape, and then go through the five things that R users need to know before developing their first Shiny for Python application.

What are Quarto shortcode extensions? Think of them as powerful little programs you can run in your Quarto docs. I won't show you how to build a shortcode extension during this talk but rather I'm going to take you on a trip across this new ecosystem of shortcode extensions that people have already written. For example, I'll introduce you to the `fancy-text` extension for outputting nicely-formatted versions of fancy strings such as LaTeX and BibTeX; you'll learn all about the `fontawesome`, `lordicon`, `academicons`, `material-icons`, and `bsicons` shortcode extensions that let you add all sorts of icons. This is only a sampling of the shortcode extensions I will present, there will be many other inspiring examples as well.

Data visualizations are a key tool in the process of sharing insights with stakeholders in ways that make impact and are easily understood. Animation and interactivity bring in a key component of analysis that enables you to derive insights that a plot with two dimensions doesn’t otherwise allow. Sometimes it can tell a powerful story and sometimes it can just add some flair to an otherwise low-key story. Plotly, Crosstalk and Highcharter are powerful packages that allow interactivity right from within your Quarto markdown report, without the need to create an entire application. Easily have two separate plots interact with each other or add drill down capabilities within a single plot, whichever adds to the story you are trying to tell.

REGENERATE is real-time decision support tool for Mayo Clinic's ICU (Intensive Care Unit). Its main aim is assure that patients with respiratory issues receive appropriate care once they discharge from the ICU. The support tool was developed, hosted, and supported entirely with R and Posit Solutions. It comprises of a machine learning model (xgboost) developed with reproducible tools such targets, renv, and git. Predictions for Mayo Clinic's ICU patients is made in real-time with the aid of vetiver and data streams supported by httr2 and DBI packages. Finally, physicians interact with a front-end UI developed in shiny and hosted on a Posit Connect server that provides recommendations to prevent patient readmission.

Every year, government agencies release large amounts of data on schools and colleges, but this information is scattered across various websites and often difficult to use. To make these data more accessible, the Urban Institute built the Education Data Portal, a freely available one-stop-shop for harmonized data and metadata for all major federal education datasets. We've also built tools around the portal including an R package, a (forthcoming) Python package, a Stata package, and an interface to generate R, Python, Stata, and JavaScript code. In this talk, we’ll demonstrate how these tools work and share lessons we’ve learned about making data accessible to users with varying technical skills and preferred programming languages.

The Cook County Assessor's Office (CCAO) determines the current market value of properties for the purpose of property taxation. Since 2020, the CCAO has used R, Tidymodels, and LightGBM to build predictive models that value Cook County's 1.5 million residential properties, which are collectively worth over $400B. These predictive models are open-source, easily replicable, and have significantly improved valuation accuracy and equity over time. Join CCAO Director of Data Science Dan Snow as he walks through the CCAO's modeling process, shares lessons learned, and offers a sneak peek at changes planned for the 2024 reassessment of Chicago.

Our talk will be about how our adoption of Quarto for document creation has transformed the collaborative workflow in a higher education analytics office. Historically, our content experts wrote in Word documents and our data analysts ran statistical analyses and made graphs in R. Specialization in different software tools created challenges for producing collaborative analytic reports, but Quarto has solved this problem. We will describe how we use Quarto for writing and editing text, embedding statistical analysis, and producing reports with a standard style in multiple formats. Speaker topics: Wyl Schuth: writing, editing, and html/markdown formatting Michael Zenz: embedding R/Python code Susan McMillan: data visualization/reporting

Recently webR v0.1.0 was released, and users have begun building new interactive experiences with R on the web. In this talk, I'll about webR's TypeScript library and the kind of things that it can do. The library aims to allow users to interact with the R environment directly from JavaScript, which allows for manipulation tricks that when not seen before can seem like magic. I'll begin by describing how to move objects from R to JS and back, and talk about the technology behind the proxy objects that make this possible. I'll continue showing advanced manipulation, such as invoking R functions from JS and vice-versa. Finally I will describe how messages are sent on webR's communication channel, building up to a demo of a "Shinylive for R".

Large language models (LLMs), such as ChatGPT, have shown potential to transform how we code. As an R package developer, I have contributed to the creation of two packages -- gptstudio and gpttools -- specifically designed to incorporate LLMs into R workflows within the RStudio environment. The integration of ChatGPT allows users to efficiently add code comments, debug scripts, and address complex coding challenges directly from RStudio. With text embedding and semantic search, we can teach ChatGPT new tricks, resulting in more precise and context-aware responses. This talk will delve into hands-on examples to showcase practical application of these models, as well as offer my perspective as recent entrant into public package development.