Initial commit

.gitignore

@@ -37,3 +37,7 @@ vignettes/*.pdf
 
 # R Environment Variables
 .Renviron
+.DS_Store
+.Rprofile
+Small-Multiples.Rproj
+renv/

Part1--Annotated-Script.Rmd

@@ -0,0 +1,254 @@
+---
+title: "Using Small Multiples to visualise the effect of the coronavirus pandemic"
+subtitle: "Part 1: [DRAFT] script for a data visualisation tutorial"
+author: "Alex Reppel"
+date: "30 August 2020"
+bibliography: references.bib
+output: github_document
+---
+
+
+# Personal introduction
+
+My name is Alex Reppel and I am a Reader in Marketing at Royal Holloway, University of London. My interest in data visualisations is **twofold**.
+
+Firstly, **on a practical level**, exploring numerical data visually helps me **identify patterns and relationships*. So, in that sense, visualisations make data accessible. Data visualisations also make it easier for me to communicate and discuss what I am seeing in the data with others.
+
+Secondly, **on a more conceptual level**, I am interested in the **aesthetic properties** and underlying **design principles** of data visualisations. Of course, there is some **debate** -- as there always is -- about principles in general and what does and what does not constitute "best practice".
+
+While some authors tend to be more prescriptive than others, I am **generally intrigued** by these discussions because, for the observer, **there is almost always something to be learned** when people disagree.
+
+
+# Topic introduction
+
+I have chosen a chart that isn't really one of the basic chart types we are familiar with. Instead, it is an **arrangement of charts**.
+
+It's called a **Small Multiple** and it's essentially a grid of charts of the same kind.
+
+------------
+
+> *Switch to screen:*
+> https://www.theguardian.com/world/2020/jul/30/covid-19-england-had-highest-excess-death-levels-in-europe-by-end-of-may
+
+Here, we see an example of a small multiple that we will explore in more detail later on. It clearly shows how charts of the same kind are arranged in a grid.
+
+- Although other labels exist, such as "grid chart" or "panel chart",
+- I personally prefer the term small multiple because it **describes the arrangement of charts really well**.
+
+We are essentially taking a **small** version of a chart and **multiplying** it a number of times based on one or two other variables.
+
+These variables are often categorical, such as the names of countries, as shown in this example.
+
+The result, then, is a grid of panels that each show a chart of the same kind, one for each country.
+
+------------
+
+> *Switch back to speaker:*
+
+Small multiples are an **intriguing concept** and **my objectives** for introducing it today are ...
+
+1. First, to expand a little bit more on **the concept itself** and its **benefits** ... 
+2. And second, to **illustrate** small multiples by using examples from the beginning of the 2020 coronavirus pandemic ... 
+
+
+# Chart type introduction
+
+We have seen before that small multiples are essentially **a grid of charts** of the same kind. ...
+
+Edward Tufte [-@Tufte:2001aa, p. 170], whom I will probably refer to quite a few times throughout this video, uses the **analogy** of a movie where charts within the grid resemble movie frames.
+
+This arrangement leads to what Tufte calls "**an economy of perception**" [@Tufte:1990aa, p. 29] where -- and I quote --
+
+    once viewers decode and comprehend the design for one slice of data
+    they have familiar access to data in all the other slices.
+
+Put differently, the "economy of perception" that Tufte associates with small multiples is the result of **us being able to comprehend the entire grid** once we have processed -- decoded -- **one element of it**.
+
+...
+
+Small multiples come in very **many forms**.
+
+They could be drawings, photographs, collages, or indeed charts.
+
+They all have in common that they, to quote Tufte [-@Tufte:1990aa, p. 68] again,
+
+    reveal, all at once,
+    a scope of alternatives,
+    a range of options
+
+with **each option** being shown in one chart within the grid.
+
+**Kieran Healy** [-@Healy:2019aa, p. 76], who wrote an excellent introduction on how to create data visualisations, including small multiples, describes them as
+
+    a powerful technique ...
+    that allows a lot of information
+    to be presented compactly
+    and in a consistently comparable way 
+
+Drawing these descriptions together, we can conclude that
+
+This **arrangement of charts** allows us to present **a lot of information** and **compare it consistently** and **efficiently**.
+
+...
+
+**Let's look at a few examples.**
+
+
+# Case studies
+
+From the beginning of the coronavirus pandemic in 2020, small multiples were used to compare outbreaks across countries.
+
+
+## Example #1: BBC
+
+------------
+
+> *Switch to screen:*
+> https://www.bbc.co.uk/news/world-51235105
+
+The example you see here is from the BBC website.
+
+This graph shows the number of confirmed cases for **six European countries** between March and August 2020.
+
+Daily numbers are shown as light blue bars, with a thicker line indicating the average number of cases over the previous seven days.
+
+We can see that **the vertical axis** -- or *Y-axis* -- is **fixed**. This means that the same values are used to scale each chart.
+
+In this example, the vertical axis for each of the six charts goes from 0 to 20,000. ...
+
+This makes it easy to **compare changes across these six countries**.
+
+We can see, for example, that cases in Spain and France are rising fast, and that the number for Belgium is considerably lower compared to those for the other five countries.
+
+...
+
+This is not surprising because what we are seeing are **absolute numbers of cases**, not the number of cases relative to the population of each country.
+
+This is an important consideration, given that the population of the UK is almost six times larger than the population of Belgium, and Germany more than seven times larger.
+
+...
+
+One option to compare the numbers for all six countries would be to show cases relative to the size of the population, such as in "Number of cases per 100,000 people".
+
+But even without this information, this small multiple from the BBC **clearly shows an increase in numbers for some countries**, which is *insightful* and does encourage us to **ask further questions**, such as whether the increase in cases **is distributed evenly within these countries** or limited to so called "hot spots".
+
+As a result, this visualisation **creates awareness**, which in itself is an important contribution data visualisations can make.
+
+
+## "Three ways to measure UK deaths"
+
+------------
+
+> *Switch to speaker:*
+
+Another way to analyse the devastating impact this pandemic had during the first half of 2020, is to look not only at the number of cases, but also at **the number of casualties**.
+
+------------
+
+> *Switch to screen:*
+> https://www.bbc.co.uk/news/uk-51768274
+
+Here, we see a diagram -- again from the BBC website -- that explains the different ways in which UK deaths related to coronavirus can be measured.
+
+...
+
+The third way -- shown here at the bottom -- is to **compare** the number of cases over a given period of time with the number of cases **we would expect for that period** based on the average over the previous five years.
+
+
+## Example #2: The Guardian
+
+> *Show new screen:*
+> https://www.theguardian.com/world/2020/jul/30/covid-19-england-had-highest-excess-death-levels-in-europe-by-end-of-may
+
+This way of measuring mortality -- often referred to as "**excess deaths**" or "**excess mortality**" -- is shown in this example of a small multiple from The Guardian.
+
+This is an interactive graph that reveals additional detail when we move over certain areas of the graph.
+
+Interactivity invites exploration and engagement with the underlying data.
+
+...
+
+In terms of visual representation, we see a grey dotted line and a red area.
+
+The grey dotted line indicates the average number of deaths that have occurred over the past five years. Typically, we would expect numbers for 2020 to be close to that line.
+
+The difference between the actual numbers for 2020 and the expected numbers based on a five year average is shown in red.
+
+This red area highlights "excess mortality". The related technical term for this feature is "preattentive pop-out", which Healy [-@Healy:2019aa, pp. 19-20] describes as
+
+    mak[ing] some things on a data graphic easier to see or find than others.
+
+...
+
+Contrary to the previous example from the BBC website, the vertical axis -- the **y-axis** -- is **not fixed** and shows **different numbers** for each nation. This is referred to as a **free scale**.
+
+Hadley Wickham [-@Wickham:2016aa, p. 151], from whom we will hear much more when we create our own chart, says that
+
+    Fixed scales make it easier to see patterns across panels;
+    free scales make it easier to see patterns within panels.
+
+Using a free scale for the vertical axis means that it is **slightly more difficult** to compare the **total number of cases** across the four UK nations, but **easier** to compare **their development over time** for each country.
+
+We see, for example, that the **slope of the curve for Scotland** differs from that of the other three nations, where we see another brief increase around week 18.
+
+
+## Example #3: Financial Times
+
+> *Show new screen:*
+> https://www.ft.com/content/a26fbf7e-48f8-11ea-aeb3-955839e06441
+
+The last example I would like to show you is a small multiple developed by John Burn-Murdock for the Financial Times.
+
+This chart allows us to compare excess deaths **across many countries** at the beginning of the 2020 outbreak.
+
+Compared to the previous two examples, this chart is considerably **more complex**.
+
+But when one takes the time to explore it, the "**economy of perception**" that Tufte associates with small multiples becomes obvious.
+
+...
+
+Similar to the previous example from The Guardian, the vertical axis -- the y-axis -- is **not fixed**, which, again, makes it easier to **compare the shape** of each line graph.
+
+This version also takes into account **differences in terms of data availability**. For example, we can see that the latest data available for the UK is June 26th, while the latest data available for Italy is April 29th.
+
+Overall, this example of a small multiple manages to put **a lot of information** into a single graph, while still being readable.
+
+# Conclusion
+
+------------
+
+> *Switch to speaker:*
+
+This last example from the Financial Times demonstrates very clearly the "**economy of perception**" that Tufte associates with small multiples.
+
+...
+
+It demonstrates -- in an elegant way -- that this form of visualisation is particularly useful when we want to present a lot of information and to compare that information consistently and efficiently.
+
+...
+
+In a second part in this series on small multiples, I am going to demonstrate how you can create a graph using R and a library called **ggplot2** [@Wickham:2016aa].
+
+The example graph we will create shows the 2020 mortality rates for England and Wales across different age groups. 
+
+------------
+
+> *Switch to graph:*
+
+![Example graph.](_export/age-groups-free-y-axis.pdf){width=100%}
+
+As you can see, the example we will be developing contains a number of customizations that were inspired by the small multiple developed by John Burn-Murdock for the Financial Times.
+
+The **ggplot2** library makes the creation of a graph easy, and the customisations based on an existing example adds a little extra challenge to the task. But, as we will see, even that can be accomplished using **ggplot2**.
+
+------------
+
+> *Switch to speaker:*
+
+Thank you very much for watching and I hope to see you soon for the second video in this series.
+
+------------
+
+# References
+
+