FFF: Ad hoc but artisanal
Hey loyal readers,
This is the part of the newsletter where I comment on the embarrassingly long interval between newsletters and make some vague excuses (or promise not to make excuses). In my February newsletter (gasp), the excitement about ChatGPT and Generative AI was still fresh and gave me a reason to write again. Ironically, the firehose of news about Generative AI has made it challenging to find interesting tidbits about ‘R and analytics in the public sector and the world of consulting’.
I did find time to share on LinkedIn what I have been doing with Generative AI (and some R and SQL) at Nous Group, but I didn’t judge that to be FFF-worthy.
In this {interval}’s newsletter, we have a mixed bag of links that cover the evergreen topic of ‘what’s wrong with pie charts’, a tool for prototyping Shiny apps, a new blog with some cool graphs on labour market data, small area population forecasting and some useful R packages for common problems of messy data.
1 - What’s wrong with pie charts?
This substack post from the data visualisation superstar Yan Holtz is a gentle take on the question of pie charts. In my opinion, this is a topic where it is easy for people into ‘dataviz’ to be snobby and sneering in a way that alienates others by not explaining their steadfast belief that ‘pie charts are bad’.
This is a useful post to bookmark and share because it gently explains the shortcomings of pie charts, notes where they are perfectly fine and also has a cool dynamic visualisation that transitions between a pie chart and bar chart.
If you aren’t already familiar with Yan’s work, check out his websites here, they include the fantastic Data to Viz and R Graph Gallery.
2 - A new blog on economic insights from labour market data (with some great graphs)
Big fan of the FFF - Matt Cowgill - is back blogging again. He is now sharing his economic insights on the labour market on the SEEK website. Interesting insights, but also some terrific graphs (made with ggplot2 and various ggplot2 extensions) in only a few lines of code…
3 - ShinyUiEditor, a drag-and-drop interface for building Shiny apps - is out of ‘alpha’
Posit (formerly RStudio) have announced that shinyuieditor is out of 'alpha'. shinyuieditor is an R package that helps you 'build shiny application UIs by dragging and dropping. It generates clean and proper code as you build.
Possibly more excitingly, Posit have made available a live demo app that you (or your less technical colleagues) can use to prototype UI (and generate code).
4 - New population projections for ‘small areas’
Some colleagues of mine recently attended a great workshop that was squarely at the intersection of analytics, public sector, academia and consulting. Presented by Tom Wilson, Irina Grossman and others, the workshop focused on techniques for small area population forecasts. Small area population forecasts (e.g. Statistical Area 3) are widely used by government and business for planning, budgeting and policy purposes but they have received relatively little research attention – unlike national-level forecasting.
The slides from the workshop are on github and the researchers have made available the dataset containing population projections by sex and age group for SA3 areas of Australia. The data are presented in both Excel and csv formats.
5 - Useful tools for working with messy data
Finally, to wrap up, here are some relatively new R packages that try and help solve the problems of working with messy data problems:
{tidygeocoder} - A unified high-level interface for a selection of supported geocoding services.
{zoomerjoin} - an R package that empowers you to fuzzy-join massive datasets rapidly, and with little memory consumption.
I came across these in the always useful and always friendly R Users Network for Australian Public Policy Slack Group. Unfortunately, the free version of Slack deletes messages after a while so I am posting them here for future reference.