Where and When

Meeting will be held 2025-11-19 17:30-19:00 in Lecture Theatre 8 of The Diamond.

Sign up via Meetup although remote attendance is possible via Google Meet.

Abstract

Access to healthcare-professional administered medicines in England varies significantly across the country, resulting in inequitable health outcomes. Our interactive web platform identifies optimal new care site locations. It uses a greedy search method outlined in Schneider & Smith (2020) to rank candidate sites that maximize a range of user-selected objective functions such as: travel distance, travel time, adherence or health outcomes. Users can customize input parameters like travel cost and CO₂/km, and estimate the impact of adding new sites on a range of outcomes including patient travel time, adherence, health outcomes, CO₂ emissions, and out-of-pocket costs. The platform uses data on the socioeconomic and demographic structure of each Lower Super Output Area, a small community of around 1500 individuals, in England. It combines this with information on the prevalence of disease, and the relationship between geographic access and outcomes. It allows users to assess the impact of creating one or more of the sites identified on national and regional outcomes, and to assess the impact on inequalities. The platform is in beta stage, and we are interested in feedback from the research community on how to build upon this work going forward.

Bios

Oliver Dolin is a data scientist at Dark Peak Analytics, where he builds web interfaces for health economic models in R Shiny. He has contributed to R packages for Health Economics and Outcomes Research (HEOR) tasks (assertHE, and an internal package for network meta-analysis plot visualizations). He holds an MSc in Health Data Science from London School of Hygiene & Tropical Medicine and a BA in Experimental Psychology from the University of Oxford.

Robert Smith is a Director of Health Economics at Dark Peak Analytics. He leads the development of health economic models, web-platforms and software for Health Economic Evaluation. He is co-director of the R for Health Technology Assessment consortium. Rob has previously advised on projects for the WHO, CGD, 10 Downing St, UKHSA and most of the top 10 pharma companies. He has a PhD in Public Health Economics & Decision Science from the University of Sheffield.

Where and When

Meeting will be held 2025-03-13 12:30-13:30 in Room 315 of The Diamond.

Sign up via Meetup although remote attendance is possible via Google Meet.

Charlotte Head - {devtools}

Want to make your R life easier? This talk introduces how to build R packages (it’s much easier than you think!) and how they can organise your code. Using devtools, we’ll build a simple package (by getting some ASCII animals to compliment you using the R packages praise and cowsay!). We’ll be covering the essentials: structure, functions, and documentation. By the end, you’ll be ready to create your own packages. Bring your laptop if you’d like to follow along (the download prerequisites for codealongs are the devtools, praise and cowsay packages, as well as Rtools (Windows) or XCode (Mac)).

Neil Lawrence - SITAR : Longitudinal multilevel modelling

Super Imposition by Translation And Rotation (SITAR) is an advanced statistical modelling that uses longitudinal multilevel modelling to appropriately summarise time series data, particularly useful for modelling the growth of children. Advantages include application to data that is measured at variable frequencies, as well as relatively simple model fit statistics that help the selection of hyperparameters and the avoidance of overfitting. However, the use of terms such as ‘random effects’, akin to terms like ‘missing at random’ when considering challenges around missing data, create barriers when explaining these models to clinicians. Technical challenges such as the impact of time varying covariates also make robust estimation of the impact of treatment changes within patients difficult to define. This talk will show the application of SITAR modelling to cohort and registry data and some of the insights that have been gained, and hopefully inspire conversation about how such modelling can move from being purely descriptive, to predictive and suitable to be incorporated into clinical decision support tools on the front line of clinical practice.

Schedule

Preparation

To participate you will need a GitHub and/or a GitLab account and to register between 2024-09-23 and 2024-10-31 on the Hacktoberfest site.

Learn More

There are a bunch of resources available at Resources for Beginners if you are unfamiliar with Git or GitHub/GitLab but these aren’t essential as we are aiming to cover the basics and help you over the course of these events.

If you are interested in making contributions to R core - the code and documentation that gets distributed as base R - check out the information from R Contributor Site.

Topics Covered

A few links to material from some of the talks given…

• rplaywright - Erika Siregar - Erika has developed a wrapper for the playwright testing framework and is keen to receive feedback from those who are able to test it.

• Detecting Changes in Population Trends - Emma Pritchard

• Sophie Lee talked about using Quarto to create accessible, engaging teaching material.

• BITSurv R Package for survival analysis - Ben

• Render R Markdown with GitHub Actions - Simon Rolph

• No Data, No Worries : Fabricatr

Erika Siregar also floated the idea of establishing a local Sheffield R Ladies chapter so if anyone is keen and would like to be involved in that please reach out and contact Erika directly.

Profiling and optimising your R code

Nobody likes slow code, luckily most programming languages provide you with tools to help you analyse and enhance your code’s performance and R is no exception.

In this session we’ll cover some of the tools and strategies available in the R ecosystem to optimise your code. We’ll start by giving a brief introduction to profiling, and then we’ll use the {profvis} to profile some code to determine which parts are taking up the most execution time. Then we’ll move on to discuss potential strategies that reduce execution time, and we’ll use the {bench} to compare these different strategies against one another to find the best performing solution.

R Strategies for Health Data Pipelines

In this session, we delve into the intricate world of health data. Our discussion encompasses crucial aspects such as data packaging, ensuring data integrity, and handling sensitive health data securely.

Navigating the complexities of health data, we address the challenges associated with the arrival of unforeseen data volumes. With a focus on scalability, we examine techniques to handle unexpectedly large data sets using packages like {arrow} and {DuckDB} and {dplyr}.

Reprex

Its possible that the section of code you are having trouble with is part of a larger workflow. To make it easier for people to understand the problem and offer assistance it would be beneficial if you are able to abstract out the problem to what is know as a Minimal “Reproducible Example” often shortened to the portmanteau “reprex”. There is advice on How to create a Minimal, Reproducible Example on Stack Overflow and there is also the R package {reprex} which helps “prepare reproducible example code via the clipboard”.

Data Security and Sensitivity

If your problem involves sensitive data then please be mindful of this. See the University of Sheffield Information Security Training pages, in particular the training on Protecting Research Data, Protecting Personal Data and Protecting Information. If you are not a student or member of staff at The University of Sheffield please refer to your institution or employers guidelines on data security.

Dive into {targets}

Delve into the dynamic world of the {targets} R package—a powerful tool designed to streamline and enhance your data science and analysis workflows.

targets is not just another package; it’s a game-changer for reproducibility, efficiency, and scalability in your R projects. Developed to manage complex dependency graphs, targets automates the execution of your analyses, making it easier to track changes.

In this session we’ll introduce the targets package, discuss why it’s useful and explore the core features of the targets package by defining a basic pipeline.

Robert Smith

Living HTA: Automating Health Economic Evaluation with R

Background: Requiring access to sensitive data can be a significant obstacle for the development of health models in the Health Economics & Outcomes Research (HEOR) setting. We demonstrate how health economic evaluation can be conducted with minimal transfer of data between parties, while automating reporting as new information becomes available.

Methods: We developed an automated analysis and reporting pipeline for health economic modelling and made the source code openly available on a GitHub repository. The pipeline consists of three parts: An economic model is constructed by the consultant using pseudo data. On the data-owner side, an application programming interface (API) is hosted on a server. This API hosts all sensitive data, so that data does not have to be provided to the consultant. An automated workflow is created, which calls the API, retrieves results, and generates a report.

Results : The application of modern data science tools and practices allows analyses of data without the need for direct access – negating the need to send sensitive data. In addition, the entire workflow can be largely automated: the analysis can be scheduled to run at defined time points (e.g. monthly), or when triggered by an event (e.g. an update to the underlying data or model code); results can be generated automatically and then be exported into a report. Documents no longer need to be revised manually.

Conclusions: This example demonstrates that it is possible, within a HEOR setting, to separate the health economic model from the data, and automate the main steps of the analysis pipeline.

Wael Mohammed

Packaging cost-effectiveness models in R: A Tutorial

Background: The use of programming languages such as R in health economics and decision science is increasing, and brings numerous benefits including increasing model development efficiency, improving transparency, and reducing human error. However, there is limited guidance on how to best develop models using R. So far, no clear consensus has emerged.

Methods: We present the advantages of creating health economic models as R packages - structured collections of functions, data sets, tests, and documentation. Assuming an intermediate understanding of R, we provide a tutorial to demonstrate how to construct a basic R package for health economic evaluation. All source code used in or referenced by this paper is available under an open-source licence.

Case Study: We use the Sick Sicker Model as a case study applying the steps from the tutorial to standardise model development, documentation and aid review. This can improve the distribution of code, thereby streamlining model development, and improving methods in health economic evaluation.

Conclusion: R packages offer a valuable framework for enhancing the quality and transparency of health economic evaluation models. Embracing better, more standardised software development practices, while fostering a collaborative culture, has the potential to significantly improve the quality of health economic models, and, ultimately, support better decision making in healthcare.

Summary

Join us for an exciting showcase where we’ll dive into two innovative R-Shiny projects. The first project introduces the fascinating world of mapping to investigate the dynamics of segregation among neighboring regions, a concept we fondly refer to as “social frontiers.” In the second project, we explore economic sectors with a focus on aiding South Yorkshire’s mayoral authority in their strategic growth planning endeavours.

The burning question that currently occupies our minds: How can we leverage powerful tools like Shiny to assist other institutions, including local councils and mayoral offices, in streamlining their data analysis processes and preventing unnecessary duplication of effort? Be a part of this engaging exploration and discover how Shiny can be a game-changer in data analysis for various institutions. Don’t miss this opportunity to be part of the conversation and gain valuable insights into the future of data-driven decision-making.

Introduction to Plotly

Unlock the power of interactive and engaging data visualizations with Plotly in R! In this workshop, you’ll dive into the world of Plotly, a leading data visualization library, and learn how to create stunning, interactive charts and graphs for your data analysis projects.

Grace Accad, Data Scientist in the Data Visualisation Team at The University of Sheffield, will showcase the use of Plotly to enhance your graphics.

Quarto for RevealJS Slides

The evolution of literate programming continues with Quarto which builds on the widely popular RMarkdown. Powered by Pandoc under the hood it continues to provide many popular document structures such as books, journal articles, Beamer articles and websites, but extends these further introducing new document structures such as blogs and RevealJS slide presentations. Further support is now available for more programming languages beyond R including Python, Julia and the JavaScript data analysis language Observable.

Neil Shephard is a Research Software Engineer within the Department of Computer Science, will guide you through how to use a GitHub Template for Quarto RevealJS slides to create and publish them to GitHub Pages.

You can view the slides that were presented here and if you wanted to try making slides yourself you can use the Quarto RevealJS Template that was demoed. The talk/slides cover the material that is on the front README.md. If you find any mistakes please do report an issue.

Speaker Bio

Dan is a researcher in the Sheffield Methods Institute working on a diverse bunch of projects, from modelling whether houses can see wind turbines or green space to making 3D prints of Census data to making monsters using a Kinect motion sensor. He also occasionally runs day workshops on data wrangling and visualisation in R.

Anna is a Research Software Engineer at the University of Sheffield, helping researchers do more with their code and data.

Managing projects with workflow

In this session I’ll do a walk through of using package workflowr to organise and document projects including some best practice tips and tricks.

Anna Krystalli has a background in marine macroecology. Now she works as a Research Software Engineer at the University of Sheffield, helping researchers do more with their code and data.

Introduction to DiagrammeR

In this session I’ll give a brief introduction to DiagrammeR, a package for creating and visualising network graphs and diagrams, with built-in support for a range of formats. I’ll demonstrate how to create a data-driven network diagram using data on trade in endangered species.

Tamora James is a PhD researcher in the Department of Animal and Plant Sciences at the University of Sheffield. Her work involves modelling populations of vulnerable bird species to determine effective conservation interventions.

Sheffield R Users Group needs you

We rely on members of the Sheffield R community to speak at our meetups and try to offer a mix of “beginners” and “expert” level talks. If you’d like to share a 15 minute informal slide-talk on a topic of your choosing (broadly ‘something cool I’ve done related to R that others would benefit from hearing about’) then we’d love to hear from you!

Let us know by filling in our speaker sign up form.

Schedule for the rest of the month

• Tu Oct 10 5:00 pm - 8:00 pm University of Sheffield, Diamond Meeting room 1

• Tu Oct 17 5:30 pm - 8:30 pm University of Sheffield, Information Commons Collaboratory 1 ‐ 1.26

• Tu Oct 24 5:30 pm - 8:30 pm University of Sheffield, Diamond Meeting room 1

• Tu Oct 31 5:30 pm - 8:30 pm University of Sheffield, Diamond Meeting room 1 - Halloween Hacktoberfest!

It’s Hacktoberfest over on GitHub! a global celebration of open source intended to promote participation in open source projects.

So this month we’re taking a break from talks and booked some group hackspaces at the University throughout the month to get involved in some open source projects requiring some #rstats help.

Anyone can benefit from getting involved, regardless of coding skill. At the very least, you’ll learn more about open source and how to contribute through online collaborative coding platforms like GitHub. But you never know, you might end up stumbling across the perfect project for you and ending up joining a community.

You can choose to contribute as little or much as you like, from correcting typos, contributing ideas to discussions to building features! Make at least 4 contributions (pull requests) to any GitHub-hosted project between October 1 and October 31 to earn a limited edition Hacktoberfest T-shirt!

At today’s meetup

We’ll take a look at Hacktoberfest R projects on GitHub (many rOpenSci packages requiring help!) and discuss how to use GitHub to contribute to projects e.g. how to make a pull request, best practice for contributing code etc.

Thermal images in R

Rebecca will introduce the ThermImage package and demonstrate its use in extracting and processing thermal images in R. She’ll talk about the challenges associated with the structure of thermal data and the kinds of questions that ecologists can ask with such data, such as capturing the extent to which animals can move locally to track their preferred temperatures. A single thermal image produces 19,200 distinct temperature measurements, so optimising the extraction -> processing -> plotting -> analysis workflow is essential.

Object oriented systems in R

Object Oriented programming is commonplace in many languages. It is a useful way to organise code. R has some special flavours of Object Oriented programming. Understanding these will help you understand R better and write better software.

We hope to see you there!

The SheffieldR team

Literate programming with Rmarkdown

In this session we’ll explore the various options and strategies for literate programming and dynamic report generation in R and Rmarkdown. This session will make use of the literate programming packages knitr and Rmarkdown.

Writing a decision tree package for R

Decision trees are a class of model widely used in health economic modelling. A tree (in the mathematical sense) has probabilities associated with branches leaving a node to describe the chances of a set of potential outcomes. In health economics, costs and health consequences are also associated with outcomes. By way of example, clinical diagnostic algorithms based on a sequence of symptom screens and diagnostic tests map particularly naturally into this framework. There are proprietary packages to develop, visualize and analyse such models, but nothing open-source and user-friendly. We describe an attempt to develop a package with a simple syntax for decision tree models in R based on the DOT language, which allows visualization, as well as considerable flexibility in analysis approach.

We hope to see you there!

The SheffieldR team

Using C++ to speed up R code

R is a very handy tool for data analysis, but it can be computationally quite slow. When analysing large data sets, or using complex models, it can be desirable to speed things up. For this reason, it has become increasingly common to combine R code with C++ code. C++ is a compiled programming language, which makes it much faster than R in some cases. I will first talk about Rcpp, an R package which makes it very simple to call C++ functions from an R script. Then, I will talk about the Template Model Builder (TMB), an R package which uses automatic differentiation in C++ to speed up the numerical evaluation of a function (typically, a likelihood function), thus making techniques such as maximum likelihood estimation very fast. I will show that these do not require a strong programming background, and that it is quite easy for someone familiar with R to use Rcpp and TMB. I will compare the three approaches with examples of code.

We hope to see you there!

The SheffieldR team