schnappi-wkl/dataviz-art-skill

Data visualization course at MPIDR

Dataviz – the Art/Skill Cocktail

In the ever-growing universe of dry academic texts, impressive and efficient graphics are quite rare. Driven by widespread software legacy issues and mostly outdated limitations imposed by traditional scientific publishers, researchers often consider producing high quality graphics as a peripheral optional task – “if time allows” (spoiler: it won’t). Yet, modern tools place data visualization in the focus of research workflows when it comes to conveying the results. Hence, the ability to turn a large dataset into an insightful visualization is an increasingly valuable skill in academia.

The course aims to empower the participants with the flexibility that the R+tidyverse framework gives to visualize data (the practical examples mostly use demographic data). The course covers some aspects of data visualization theory and best/worst practice examples, but it's primarily practice oriented including live coding sessions and short lecture/showcase parts.

Practical coding sessions start from basic introduction to tidy data manipulation and ggplot2 basics. Next, practical examples cover the creation of certain most useful types of plots. Important data visualization choices and caveats are discussed along the way. Special attention is devoted to producing geographical maps, which are no longer the luxury of professional cartographers but have turned, with the help of R, into yet another data visualization type. Going beyond ggplot2, the course presents an introduction to interactive data visualization.

Course Twitter account: @DatavizArtSkill

Video lectures: https://bit.ly/dataviz-art-skill

Prerequisites

• R
• Rstudio
• Git (help page)
• R {packages}
• Basic familiarity with R, consider RStudio Primers

Outline

Day 1: BASICS

• Basic dataviz principles (slides)
• Impressive dataviz showcases
• tidtverse: Tidy approach to data

Day 2: TUNE-UP

• ggplot2 basics
• Colors in dataviz
• Themes and fonts

Day 3: TOOLBOX

• Useful types of plots
• Dotplot – the most neglected and powerful type of dataviz
• Heatmap, equality-line, ggridges, treemap, ternary plots

Day 4: MAPS

• The basics of map projections
• With geom_sf maps become yet another type of dataviz
• Useful spatial processing tricks with rmapshaper
• biscale maps

Day 5: ROCK

• gganimate: breathe life into your plots
• Affordable interactivity with plotly
• The logic of user-input apps with self-written functions
• Cursory intro to shiny

Useful links

• Garrick Aden-Buie's Gentle Guide to the Grammar of Graphics
• Oscar Baruffa's Big Book of R
• Hadley Wickham's R For Data Science
• Neal Grantham's TidyTuesday Rocks app
• Nathaniel Smith's talk on the principles of viridis palettes
• John Burn-Murdoch's RStudio 2021 talk
• Hans Rosling: sample talk; Factfulness; gapminder.org
• Thread on excess deaths plots
• Thread on log transformation of the ratios
• Post on efficient RStudio layout
• Replication materials for our recent PNAS paper
• Boxplot + jitter example: tweet
• Jonas Schoeley's HMD explorer app
• US names shiny app: repository; tweet; shiny app