Virtual useR! 2024

I presume that, as R developers, we’ve heard of or perhaps used shiny package to develop web applications, and that, we can attest to the fact that it has emerged as a popular framework for developing dashboards in R. These shiny dashboards provide a dynamic platform for data exploration, analysis, and dissemination, making them invaluable tools for researchers, analysts, and decision-makers. However, as the volume and complexity of data grow, maintaining and updating these dashboards manually can become a time-consuming and error-prone process. Automating updates to shiny dashboards deployed on shiny server offers a solution to this challenge, enabling developers to streamline the deployment pipeline, enhance efficiency, and ensure the timely dissemination of accurate information. This session will be a hands-on tutorial that will take you through the process of setting up the workflow that works in such a way that, a push event to the main branch of a GitHub repository triggers a webhook that in turn sends a payload to an API which also triggers a bash script to do a git pull from the same repository.

Tutorial Materials: https://github.com/oyogo/useR2024_tutorial_materials

Missing data are ubiquitous, pervasive, and often ignored in statistical analyses. Unfortunately, default methods such as complete case analysis may lead to biased and invalid results. This hands-on tutorial aims to equip data analysts with knowledge and skills to validly handle missing data using the popular R package {mice}. {mice} implements multiple imputation by chained equations, a flexible method for imputing (i.e. filling in) missing entries. The session will combine theoretical insights with hands-on exercises. Attendees will first learn the fundamentals of missing data theory, and then gain practical experience in addressing real-world missing data problems through guided demonstrations and exercises. Attendees are encouraged to bring their own incomplete datasets, to implement and evaluate their newfound skills. By the end of the session, attendees will be able to make informed decisions on how to validly handle missing data in their own data analysis projects.

The Piecewise Exponential Additive Mixed Model (PAMM) has gained popularity in various domains due to its ability to tackle a wide variety of survival tasks and its flexibility to model multivariate non-linear covariate effects, including time-varying effects and cumulative effects. One advantage of this model class is the ability to use different backends for estimation. However, in order to be useful in practice, their use requires pre-processing, which differs depending on the survival task at hand and post-processing (e.g. transforming estimated parameters to quantities like survival or transition probabilities). The R package pammtools (https://adibender.github.io/pammtools/) facilitates the entire modeling process. In this tutorial, we illustrate how to apply the model class in different settings, including left-truncation, recurrent events and multi-state models.

The R Project has a global and active community with members speaking different languages around the world, often with the need or preference to be able to use R a language instead of English. To support this, R Core has implemented GNU gettext helpers enabling the translation of messages, warnings, errors etc since R version 2.1.0 (April 2005). This tutorial will provide a short overview of the related history; discuss how translations are managed in base R; review the standard PO file format used by gettext; describe the traditional process for contributing patches to R Core, and then introduce Weblate, a web-based PO file editor that simplifies the translation process. By the end of the tutorial, you should be able to translate messages from base R into your natural language of choice.

The workshop will focus on different deployment designs of machine learning pipelines using R, open-source applications, and free-tier tools. We will use the US hourly demand for electricity data from the EIA API to demonstrate the deployment of a pipeline with GitHub Actions and Docker that fully automates the data refresh process and generates a forecast on a regular basis. This includes the use of open-source tools such as point-blank to monitor the health of the data and the model's success. Last but not least, we will use Quarto doc to set up the monitoring dashboard and deploy it on GitHub Pages.

In this tutorial, I’ll present an overview of our book, Causal Inference in R, freely available at r-causal.org. We’ll discuss the whole game, so to speak, of causal inference, following a few key steps: 1. Specify a causal question 2. Draw our assumptions using a causal diagram 3. Model our assumptions 4. Diagnose our models 5. Estimate the causal effect, and 6. Conduct sensitivity analysis on the effect estimate. We’ll discuss some new tools in the causal inference ecosystem, such as tipr, ggdag, propensity, halfmoon, and more, each making the act of causal inference easier and more principled.