-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathREADME.Rmd
320 lines (254 loc) · 13.8 KB
/
README.Rmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
---
title: "Recreating the New York Times COVID-19 Spiral Graph"
output: github_document
---
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE,
fig.retina = 1.5,
fig.width = 5,
fig.height = 5,
dpi = 300)
```
In a recent [opinion piece](https://www.nytimes.com/2022/01/06/opinion/omicron-covid-us.html) the New York Times included this spiral graph showing the development of confirmed COVID-19 cases in the United States since the beginning of the pandemic. This visualization has stirred some debate whether it is "the proper way" to display such data. Arguments have been exchanged why this visualization is particularly bad or why it might actually be well suitable for this use case.
## Recreating the Spiral
I will let everyone have their opinion, however, I challenged myself to recreate this plot using R and especially the [ggplot2 package](https://ggplot2.tidyverse.org/). *(Hint: I like it.)*
### Packages
Let's start with loading the R packages we will use for creating this plot.
```{r packages}
pacman::p_load("tidyverse", "ggtext", "here", "lubridate")
```
### Load and Prepare the Data
We use the COVID-19 Dataset by [Our World in Data](https://github.com/owid/covid-19-data) for the number of confirmed Coronavirus cases in the United States.
Since the original chart starts on January 1st, 2020, while the first cases in the
U.S. had been registered by end of January 2020, we add all the days from January 1st to the
first date in the data. We manage this using `tibble::add_row` and `tidyr::complete`.
(We could also insert all missing dates manually, but maybe we want to reproduce
the chart for other countries?)
We calculate the day of the year as well as the year from the date variable. The day of the year will be our x values in the plot. We will use the year to group the data in order to display the information in a cyclic manner.
```{r data, message=FALSE, cache=TRUE}
owid_url <- "https://github.com/owid/covid-19-data/blob/master/public/data/owid-covid-data.csv?raw=true"
country <- "United States"
covid <- read_csv(owid_url)
covid_cases <- covid %>%
filter(location == country) %>%
select(date, new_cases, new_cases_smoothed) %>%
arrange(date) %>%
# Add the dates before the 1st confirmed case
add_row(date = as_date("2020-01-01"), new_cases = 0, new_cases_smoothed = 0,
.before = 1) %>%
complete(date = seq(min(.$date), max(.$date), by = 1),
fill = list(new_cases = 0, new_cases_smoothed = 0)) %>%
mutate(day_of_year = yday(date),
year = year(date)
)
```
### A very basic plot
Since we want to display the data in a cyclic manner, we using a polar coordinate system with `coord_polar`. The line is built from connecting each day via `geom_segment`, which takes values for x (current day) and xend (next day)and y and yend - the latter ones will be set to an integer value (UNIX timestamp) with the function `as.POSIXct`.
```{r}
p <- covid_cases %>%
ggplot() +
geom_segment(aes(x = day_of_year, xend = day_of_year + 1,
y = as.POSIXct(date), yend = as.POSIXct(date))) +
coord_polar()
p
```
We apply `theme_void` to get rid of any theme element.
```{r}
p + theme_void()
```
### Here comes the Worm
Once we have the basic spiral, we want to encode the number of confirmed COVID-19 cases with the size of the line. For this purpose we use the `geom_ribbon` geom. For each x value, `geom_ribbon` displays a y interval defined by ymin and ymax.
But which values to use? For `x`, we use the `day_of_the_year` column. The `ymin` and `ymax` aesthetics are calculated from the UNIX timestamp and the number of confirmed COVID-19 cases. For `ymin`, we *subtract* the half of the number of cases, for `ymax`, we *add* same half of the number of cases.
Since the y values are pretty high in this plot (seconds of the year), we multiply the number of cases with a certain factor to make the ribbon visible. I have chosen 60 as a good factor.
We also add colors for the area and the outline.
```{r}
size_factor <- 60
# Colors
outline_color <- "#D97C86"
fill_color <- "#F0C0C1"
base_grey <- "grey28"
p <- covid_cases %>%
ggplot() +
# area to encode the number of cases
geom_ribbon(aes(x = day_of_year,
ymin = as.POSIXct(date) - new_cases_smoothed / 2 * size_factor,
ymax = as.POSIXct(date) + new_cases_smoothed / 2 * size_factor,
group = year),
size = 0.3, col = outline_color, fill = fill_color, show.legend = FALSE) +
# basic line
geom_segment(aes(x = day_of_year, xend = day_of_year + 1,
y = as.POSIXct(date), yend = as.POSIXct(date)),
col = base_grey, size = 0.3) +
coord_polar() +
theme_void()
p
```
And voilà, here comes the worm.
### Grid lines
In order to plot the grid lines, we need month breaks. We can calculate the breaks from the length of the months.
```{r}
month_length <- c(31, 28, 31, 30, 31, 30,
31, 31, 30, 31, 30, 31)
month_breaks <- cumsum(month_length) - 30
```
To apply the grid, we use `scale_x_continuous` and add some theme elements. We use quarters of the year as major breaks and and months as minor breaks. We label the major grid lines. (The built-in vector `month.abb` contains abbreviations of month names, but we customize the month labels so that they are in line with the NYT chart.)
```{r}
p + scale_x_continuous(minor_breaks = month_breaks,
breaks = month_breaks[c(1, 4, 7, 10)],
labels = c("Jan.", "April", "July", "Oct.")) +
theme(
plot.background = element_rect(color = NA, fill = "white"),
panel.grid.major.x = element_line(color = "grey70", size = 0.2, linetype = "dotted"),
panel.grid.minor.x = element_line(color = "grey70", size = 0.2, linetype = "dotted"),
axis.text.x = element_text(color = base_grey, size = 5, hjust = 0.5),
)
```
Notice the small breaks in the spiral with every new year? Two things are happening here: First, the scale is expanded slightly, we will disable this `ggplot2` default behavior. But secondly and more importantly, *2020 was a leap year*.
I am wondering how the team at New York Times handled it. My quick-and-dirty workaround: drop February 29th, 2020. Given that we present rolling 7-day averages this seems acceptable.
In the original plot, the spiral does not start in the center. We will add some buffer to the limits of the polar y-axis.
```{r warning=FALSE}
p <- covid_cases %>%
# 2020 is a leap year, we could drop Feb 29, 2020 for the sake of 365-day years
filter(date != as_date("2020-02-29")) %>%
group_by(year) %>%
mutate(day_of_year = row_number()) %>%
ungroup() %>%
ggplot() +
# area
geom_ribbon(aes(x = day_of_year,
ymin = as.POSIXct(date) - new_cases_smoothed / 2 * size_factor,
ymax = as.POSIXct(date) + new_cases_smoothed / 2 * size_factor,
group = year),
color = outline_color, size = 0.3, fill = fill_color, show.legend = FALSE) +
# basic line
geom_segment(aes(x = day_of_year, xend = day_of_year + 1,
y = as.POSIXct(date), yend = as.POSIXct(date)),
col = base_grey, size = 0.3) +
scale_x_continuous(minor_breaks = month_breaks,
breaks = month_breaks[c(1, 4, 7, 10)],
labels = c("Jan.", "April", "July", "Oct."),
limits = c(1, 365),
expand = c(0, 0)
) +
#' set the lower limit of the y-axis to a date before 2020
#' so that the spiral does not start in the center point
scale_y_continuous(limits = c(as.POSIXct("2019-07-01"), NA),
expand = c(0, 0)) +
coord_polar() +
theme_void() +
theme(
plot.background = element_rect(color = NA, fill = "white"),
panel.grid.major.x = element_line(color = "grey70", size = 0.2, linetype = "dotted"),
panel.grid.minor.x = element_line(color = "grey70", size = 0.2, linetype = "dotted"),
axis.text.x = element_text(color = base_grey, size = 5, hjust = 0.5)
)
p
```
### Annotations
The original chart contains a couple of annotations: the years are labelled in the spiral as a clue of the direction to read the plot, the date of the latest data in the title, and the description that 7-day averages are shown.
```{r warning=FALSE}
text_color <- rgb(18, 18, 18, maxColorValue = 255)
base_family <- "Libre Franklin Medium"
# base_family <- "Helvetica"
subtitle_date <- max(covid_cases$date) %>%
format("%b. %d, %Y")
# Annotations for the years in a list (used in annotate())
year_annotations <- list(
year = 2020:2022,
x = rep(3, 3),
y = as.POSIXct(paste(2020:2022, "01", "01", sep = "-"))
)
p <- covid_cases %>%
# 2020 is a leap year, we could drop Feb 29, 2020 for the sake of 365-day years
filter(date != as_date("2020-02-29")) %>%
group_by(year) %>%
mutate(day_of_year = row_number()) %>%
ungroup() %>%
ggplot() +
# area
geom_ribbon(aes(x = day_of_year,
ymin = as.POSIXct(date) - new_cases_smoothed / 2 * size_factor,
ymax = as.POSIXct(date) + new_cases_smoothed / 2 * size_factor,
group = year),
color = outline_color, size = 0.3, fill = fill_color, show.legend = FALSE) +
# basic line
geom_segment(aes(x = day_of_year, xend = day_of_year + 1,
y = as.POSIXct(date), yend = as.POSIXct(date)),
col = base_grey, size = 0.3) +
# annotation: 7d average
annotate("richtext",
label = "7-day<br>average",
x = 20, y = as.POSIXct("2021-08-01"),
family = base_family, size = 2, color = text_color,
label.colour = NA, fill = NA) +
annotate("segment",
x = 20, xend = 22.5,
y = as.POSIXct("2021-06-01"), yend = as.POSIXct("2021-03-15"),
color = text_color, size = 0.3) +
# annotation: years
annotate("text", label = paste0(year_annotations$year, "\u2192"), x = year_annotations$x,
y = year_annotations$y,
family = "Arial",
size = 1.5, vjust = -0.6, hjust = 0.15) +
scale_x_continuous(minor_breaks = month_breaks,
breaks = month_breaks[c(1, 4, 7, 10)],
labels = c("Jan.", "April", "July", "Oct."),
limits = c(1, 365),
expand = c(0, 0)
) +
#' set the lower limit of the y-axis to a date before 2020
#' so that the spiral does not start in the center point
scale_y_continuous(limits = c(as.POSIXct("2019-07-01"), NA),
expand = c(0, 0)) +
coord_polar() +
labs(
subtitle = subtitle_date
) +
theme_void(base_family = base_family) +
theme(
plot.background = element_rect(color = NA, fill = "white"),
panel.grid.major.x = element_line(color = "grey70", size = 0.2, linetype = "dotted"),
panel.grid.minor.x = element_line(color = "grey70", size = 0.2, linetype = "dotted"),
axis.text.x = element_text(color = base_grey, size = 5, hjust = 0.5),
text = element_text(color = text_color),
plot.subtitle = element_text(hjust = 0.5, size = 5)
)
p
```
That's pretty close!
At this point, I have refrained from reproducing the rounded text for the years. The [`geomtextpath` package](https://github.com/AllanCameron/geomtextpath) might come in handy here.
Finally, we want to add the legend. The `inset_element` function from the [`patchwork` package](https://github.com/thomasp85/patchwork) by Thomas Lin Pedersen allows to add a plot inside another one. Hence, we create a separate plot for the legend which - roughly approximately - represents the data from the plot and then used `inset_element` to add it into the spiral plot. (This is more "fake it till you make it", I am open for any suggestions to handle this programmatically.)
```{r plot-inset-legend, warning=FALSE}
library(patchwork)
p_legend <-
tibble(
cases = c(0, 150000),
ymin = c(0, -75000),
ymax = c(0, 75000),
) %>%
ggplot(aes(cases)) +
geom_ribbon(aes(ymin = size_factor * ymin, ymax = size_factor * ymax),
color = outline_color, fill = fill_color, size = 0.3) +
geom_line(aes(y = 1), color = base_grey) +
geom_text(aes(label = ifelse(cases == 0, 0, "150k cases"),
y = 1, hjust = ifelse(cases == 0, 1.5, -0.1)),
size = 2) +
coord_cartesian(xlim = c(0, 350000),
ylim = c(-as.numeric(as.POSIXct("1971-01-01")), NA),
clip = "off") +
labs(title = "New Covid-19 cases,<br>United States") +
theme_void() +
theme(plot.title = element_markdown(color = text_color,
family = "Helvetica",
face = "bold", size = 8, hjust = 0.5,
lineheight = 1.1))
ragg::agg_png(here("plots", "nyt_spiral_with-legend.png"),
res = 300, width = 1200, height = 1200 * 746/615)
p + inset_element(p_legend, left = 0.05, bottom = 0.725, right = 0.25, top = 0.95)
invisible(dev.off())
```
| Original spiral | Recreated spiral |
:-------------------------:|:-------------------------:
|  |  |
The biggest difference between original and recreated chart: The grid lines in the original chart differ in length (compare January to July). This could probably be done in a post-processing step in tools like Adobe Illustrator or Figma. I would be glad about any advice how to solve this programmatically in `ggplot2` directly, i.e. without drawing the grid lines as geoms. I could also think of editing the plot with the [`magick` package](https://github.com/ropensci/magick) for advanced image processing in R.