iiinova — where your digital self keeps evolving

Simon Smart @simonsmart.bsky.social · 09/09/2026

#rstats When you have a choice of characters as a parameter to a function, how should the choices be formatted in the docs? I have previously used option A without ever checking what the "correct" way is. 1/3

3 options for formatting:

character. Outcome type for the dataset. Either `binary` or `continuous`.
character. Outcome type for the dataset. Either "binary" or "continuous".
character. Outcome type for the dataset. Either `"binary"` or `"continuous"`.

250

Simon Smart @simonsmart.bsky.social · 18/06/2026

As a side-effect we also created a way to conduct a variety of NMAs through a simple API without using the app. Once the data is configured, you can generate some outputs or pass the data to a model function and use the result to generate more outputs. 4/5

configured_data <- setup_load("my_data.csv", outcome = "continuous") |>
setup_configure(reference_treatment = "Placebo",
effects = "random",
outcome_measure = "MD",
ranking_option = "good",
seed = 123)
freq_forest(configured_data) |>
write_plot("frequentist_forest_plot.pdf")

bayes_model(configured_data) |>
bayes_forest() |>
write_plot("bayesian_forest_plot.pdf")

120

Simon Smart @simonsmart.bsky.social · 18/06/2026

It was already one of the most impactful academic shiny apps with > 400 citations, but this update makes it reproducible using {shinyscholar}. Whatever you do in the app, you can download a Quarto document that reproduces every step or render it to produce an html report. 2/5

Example Quarto and html reports produced by the app

111

Simon Smart @simonsmart.bsky.social · 18/06/2026

Metainsight is now available on CRAN! cran.r-project.org/package=meta... It is a shiny app for conducting network meta-analysis, which in simple terms, is used by doctors to decide which of multiple treatments is best to treat your ailments. #rstats #metaanalysis #openscience 1/5

A screenshot of the app running show options in the sidebar and two forest plots displayed side by side

161

Simon Smart @simonsmart.bsky.social · 13/06/2026

Saturday meetings aren't great but can't get a much better venue to chat about #rstats

000

Simon Smart @simonsmart.bsky.social · 26/03/2026

I’m giving a seminar on 20th April about refactoring crsu.shinyapps.io/MetaInsight/ to use {shinyscholar}, producing crsu.shinyapps.io/MetaInsight_... Details and sign up: www.eventbrite.co.uk/e/crsu-webin... A shorter version is also coming to R/Medicine in May #rstats #metaanalysis #openscience

Poster for the seminar: MetaInsight v7: Making a comprehensive app for network meta analysis reproducible

002

Simon Smart @simonsmart.bsky.social · 05/03/2026

Partial matching of names in #rstats is wild. I knew the function would run, but didn't know it "works" with dataframes as well!

$f <- function(very_fast, incredibly_dangerous){ data.frame("very_fast" = very_fast, "incredibly_dangerous" = incredibly_dangerous) } df <- f(v = 1, i = 2) df$very * df$inc df$very_broken <- 3 df$very * df$inc$

341

Simon Smart @simonsmart.bsky.social · 10/02/2026

Is there a term for figuring something out that LLMs say isn't possible? Grind coding? Organic coding? In this case, rendering <text> elements in an SVG to <path> in #Rstats. (Claude cooked up a whole non-functioning package in an attempt to do it).

'<svg><text x="0" y="200"
style="font-size: 100px;">
Hello world</text></svg>' |>
charToRaw() |>
rsvg::rsvg_svg(NULL) |>
rawToChar()

010

Simon Smart @simonsmart.bsky.social · 20/10/2025

#rstats when I run devtools::test() from the Build pane I get an error that mirai isn't installed even though it is. I use r2u and have already uninstalled with apt remove and reinstalled. It runs okay from the terminal but whatever environment shinytest2 runs in can't find mirai either. Any ideas?

==> devtools::test()

ℹ Testing metainsight
Error in load_imports(path) : The package "mirai" is required.
Calls: <Anonymous> ... load_imports -> deps_check_installed -> check_installed
Execution halted

Exited with status 1.

210

Simon Smart @simonsmart.bsky.social · 21/08/2025

I've made digital art using .svgs for many years and inspired by #rtistry I've made a #rstats #rshiny app which generates them: simonsmart.shinyapps.io/savage-patte... the video shows some example outputs.

1134

Simon Smart @simonsmart.bsky.social · 21/07/2025

Myself and @statsforbios.bsky.social are running a workshop in September to introduce Disagapp disagapp.le.ac.uk for disaggregation regression. More details and sign up at forms.office.com/e/pjfbD6RvTi #rstats #episky #AcademicSky

072

Simon Smart @simonsmart.bsky.social · 29/04/2025

Being able to cancel and restart slow-running jobs in shiny with {mirai} is a bit of a game-changer #rstats Full app in alt text / gist.github.com/simon-smart8... with a method for updating progress too

$library(shiny) library(bslib) library(mirai) daemons(1) onStop(function() daemons(0)) ui <- page_fluid( p("The time is ", textOutput("current_time", inline = TRUE)), hr(), numericInput("n", "Sample size (n)", 100), numericInput("delay", "Seconds to take for plot", 5), actionButton("btn", "Go"), textOutput("progress_log"), plotOutput("plot"), ) my_func <- function(x, y, file){ writeLines(paste0("Starting..."), file) for (i in 1:y){ Sys.sleep(1) writeLines(paste0("Step ", i, " complete"), file) } writeLines(paste0("Finished"), file) runif(x) } server <- function(input, output, session) { progress_file <- tempfile() output$current_time <- renderText({ invalidateLater(1000) format(Sys.time(), "%H:%M:%S %p") }) task_var <- NULL task <- ExtendedTask$new( function(...) task_var <<- mirai(run(...), environment(), .args = list(run = my_func)) ) observeEvent(input$btn, { if (task$status() == "running"){ mirai::stop_mirai(task_var) } task$invoke(input$n, input$delay, progress_file)} ) output$plot <- renderPlot({ if (task$status() == "success"){ hist(task$result()) } }) progress_reader <- reactiveFileReader(100, session, progress_file, readFunc = function(file) { if (file.exists(file)) { readLines(file) } } ) output$progress_log <- renderPrint({ cat(paste(progress_reader(), collapse = "\n")) }) } shinyApp(ui = ui, server = server)$

1114

Simon Smart @simonsmart.bsky.social · 10/02/2025

create_template() makes a skeleton application composed of shiny modules that can optionally contain a datatable, leaflet map and display source code. Each module calls one function and suits am Rmd-first approach to development, that also minimises reactivity and opens functionality to testing.

Flowchart of the suggested workflow for developing modules, prototyped in an rmarkwdown, transferred to a function, unit tests written, develop the shiny module, reproduce the module in rmarkdown and then write end-to-end tests.

200

Simon Smart @simonsmart.bsky.social · 10/02/2025

Looking at fifty academic shiny apps published in 2023 we found that few of them had the same features and many displayed sub-optimal standards in software development. Some apps had thousands of lines of code in a single file and testing was very rare.

A graph showing the proportion of academic apps published in 2023 that cite packages, are deployed, provide instructions, provide logging, are reproducible, can be saved and loaded, are open source, include functions, are modular, are packaged and include unit and end-to-end tests.

151

Simon Smart @simonsmart.bsky.social · 10/02/2025

It shouldn’t be necessary to have read a user guide or watch a video to use an app and so shinyscholar apps provide guidance users on how to use each module as well as the scientific basis of the module.

A screenshot of the module guidance documentation inside the demonstration app.

110

Simon Smart @simonsmart.bsky.social · 10/02/2025

Once a user has conducted an analysis they can download an rmarkdown document that replicates all the steps. In addition, apps made with shinyscholar can be saved and loaded, have a flexible logging system, support asynchronous operations and attribute package developers.

A screenshot of an rmarkdown document reproducing an analysis in shinyscholar

112

Simon Smart @simonsmart.bsky.social · 10/02/2025

My first R package is now available on CRAN! Shinyscholar is a template for creating complex analytical applications that are reproducible outside the application. 🧵 cran.r-project.org/web/packages... #rstats #rshiny #openscience

A screenshot of an R shiny application created by shinyscholar, with user inputs on the left sidebar, a tab panel to move between components and a main panel featuring a map, table and results.

2187

Simon Smart @simonsmart.bsky.social · 06/12/2024

Day 6 part 1 of #adventofcode in #Rstats. I learnt from day 4 and padded the input from the start which made it a lot simpler. Barely any use of apply which makes a change from so far.

020

Simon Smart @simonsmart.bsky.social · 05/12/2024

Day 5 part 2 of #adventofcode in base #Rstats. This took forever to work out until I realised I could always put the first wrong element at the end

170

Simon Smart @simonsmart.bsky.social · 05/12/2024

Just in time for Day 5 part 1 of #adventofcode in base #Rstats. Not super pleased with this and I'm sure there is some room for improvement, but it works.

060

Simon Smart @simonsmart.bsky.social · 04/12/2024

And part 2 which was pretty simple to adapt from part 1

000

Simon Smart @simonsmart.bsky.social · 04/12/2024

Day 4 part 1 of #adventofcode in base #Rstats. This was a whole lot easier once the input was padded. I was surprised there is no function equivalent to rdrr.io/cran/terra/m... for a matrix (or is there?)

291

Simon Smart @simonsmart.bsky.social · 04/12/2024

Day 3 of #adventofcode in base #Rstats and I managed to write some regex all by myself!

090

Simon Smart @simonsmart.bsky.social · 02/12/2024

I see. I went with this so it was natural to use it for both, but your solution does prevent checking the diff unnecessarily. Do you need the double || ? Thought that was just if one could be NULL

220