Telling stories with data

James Goldie

Hi!

Climate researcher

↓

Data journalist

Storytelling fundamentals

Audience, medium and message

Audience

Do they have specialist training? (In your field, or another?)

What are their wants and worries?

Medium

Credit: MCCCRH

vs.

Credit: Abhijith S Nair/Unsplash

Medium

Where will people encounter your work?

How long will they have to consider it?

Did they seek it out?

Message

What will your readers see first?

This is called visual hierarchy.

What did you see first on this slide? What did you see second?

Why?

When all of the elements on your plot look the same, it’s really hard for people to decide what to focus on. Without that focus, it’s difficult for your plot to lead to a conclusion. Instead, you want to vary your plot elements, drawing attention to the most important things first. There are lots of ways to do this: bolding text, making it bigger, using colours, and taking advantage of natural reading direction can all help to establish visual hierarchy.

When all of the elements on your plot look the same, it’s really hard for people to decide what to focus on.

Without that focus, it’s difficult for your plot to lead to a conclusion.

Instead, you want to vary your plot elements, drawing attention to the most important things first.

There are lots of ways to do this:

bolding text
making it bigger

using striking colours…
… or softer colours

taking advantage of natural reading direction

These techniques can all help establish visual hierarchy.

Using Quarto for storytelling

What is Quarto?

The source text of a Quarto document

The visual editor’s view of a Quarto document

Example 1

Your data is electricity demand and temperature at 3 inner Melbourne sites:

github.com/jimjam-slam/talk-storytelling-r-quarto/
blob/main/data/power-temperature-threesites.csv

powerTemp = FileAttachment("data/power-temperature-threesites.csv").csv({ typed: true })

Inputs.table(powerTemp)

A basic chart

ggplot(power_temp) +
  aes(x = tmax, y = demand, colour = working) +
  geom_point(size = 2) +
  facet_wrap(vars(name)) +
  labs(
    x = "2m air temperature [°C]",
    y = "Electricity demand [MW]",
    title = "Electricity and temperature")

What stories could we tell with this data?

What improvements could we make?

Use Quarto to explore this data

R script

library(tidyverse)

ggplot(data) +
  aes() +
  # ...

R code chunk in Quarto doc

---
title: "Power and temperature"
---

Let's look at this data!

```{r}
library(tidyverse)

ggplot(data) +
  aes() +
  # ...
```

Scrollytelling with Closeread

What’s scrollytelling?

Scrollytelling involves graphics that stick to the screen and and change as the user scrolls.

ABC scrollytelling story on fake AI imagery

New York Times scrollytelling story on "Shah Jahan on a terrace, holding a pendant set with his portrait" by Chitarman

Closeread: scrollytelling with Quarto

Closeread is a Quarto extension developed by Andrew Bray
and James Goldie (me!)

(We named it after a New York Times series — the piece in the last slide was from this series!)

Closeread lets you turn a regular Quarto doc into a scrollytelling doc

How Closeread works

Step 1

A Quarto extension is like an R package — it extends Quarto’s abilities.

Install Closeread in the RStudio Terminal by running:

quarto add qmd-lab/closeread

Step 2

At the top of your Quarto doc, change
format: html → format: closeread-html

Step 3

Tell Quarto where the scrollytelling starts and stops:

:::{.cr-section}

(scrollytelling stuff goes here!)

:::

How Closeread works

Closeread has two components: stickies and triggers.

Stickies

Things you want to stick on the screen. Give them an ID starting with #cr-.

:::{#cr-myplot}

```{r}
ggplot(mydata) +
 aes(x = x, y = y) +
 geom_point()
```

:::

Triggers

These make the stickies appear. They work like citations: use the ID of the sticky with an @:

When this text scrolls past, my great figure will appear! @cr-myplot

How can scrollytelling help my communication?

Here’s a Closeread example where we build a plot up one layer at a time:

closeread.dev/
gallery/demos/
build-up-plots/

How can we use this technique with our electricity and temperature data?

Wrapup

Use Closeread, win prizes

Andrew Bray (my collaborator on Closeread) and I have teamed up with data science company Posit to run the Closeread Prize.

posit.co/blog/closeread-prize-announcement

Entries close Jan 5th
(we just extended the contest!)

Thanks! Questions?

@jamesgoldie.dev

@jimjamslam@aus.social

linkedin.com/in/jamesgoldie