library(tidyverse, warn.conflicts = FALSE, quietly = TRUE, logical.return = FALSE )
library(glue, warn.conflicts = FALSE, quietly = TRUE)
library(grDevices, warn.conflicts = FALSE, quietly = TRUE) # mostly unnecesary
billionaires_df <- read_tsv(glue("~/R_BEGINNERS_SHORT/",
"datasets_ATRIUM/billionaires_combined.tsv"),
show_col_types = FALSE)Tackling common plotting issues in ggplot2 and knitr
1 How to use this file
Open the qmd source and view it in the Visual editor. Open the html rendering nearby and compare the outcomes.
2 Common issues
Plot renders too small/narrow/low.
Axis tick labels overlap.
Axis ticks should be denser.
Texts render too small.
Overlap in text labels.
The publisher wants it black and white.
3 Setup
When you load tidyverse , it loads some of them together, definitely dplyr and ggplot.
4 Problem: Plot too small
mostly just pre-rendering and rendered OK, saved OK
default: 178 x 178 mm (or equivalent in other units)
billionaires_df %>%
ggplot() +
geom_smooth(mapping = aes(x = age,
color = sex,
y = daily_income)) +
facet_wrap(~ time)
5 Control dimensions when saving
#### Erase #| eval: false when you want this script to run on Run or Render
ggsave(glue("my_output_files", "/10_ggplotOther_plot_too_small.png"),
device = grDevices::png, units = "mm")Saving 178 x 178 mm imageThese are the dimensions of the file:
ggsave("my_output_files/10_ggplotOther_plot_too_small_height270.png",
units = "mm", height = 270, device = grDevices::png)Saving 178 x 270 mm imageTrue dimensions in the file:
or you can multiply both directions with something. Below is a very disproportionate plot just to make the effects very visible:
ggsave("my_output_files/10_ggplotOther_plot_too_small_scaleboth.png",
units = "mm", height = 178 * 1.1, width = 178 * 0.7, device = grDevices::png)
6 You want to render it better
Sometimes you need to render a document with plots for sharing. One solution could be like the one above: you generate the plots and save them. Then you insert the files as images and prevent the code from running again. The cleaner option is that you fiddle with the dimensions in the knitr chunk options.
knitr is one of the libraries that allow Quarto to render markdown documents in HTML, PDF, and doc. It pays to have an idea about the existence of knitr chunk options - be it merely as key words for AI prompts!
Below you see activated a few of them: fig.width, etc. , if you look at this file in Quarto. You will not see them in the rendered documents of any format! They belong to the start of the chunk, and the code to control the rendered width and height would look like this:
```{r}
#| fig.width: 7
#| fig.height: 10
# your code or comments follow
```The units are probably inches. The options are spelled in an older way using dots, but Help lists them with dashes instead (i.e., fig-weight instead of fig.weight. Our platform currently uses a slightly outdated version of RStudio from 2023 that ignores the new spelling (2023.06.1+524 “Mountain Hydrangea” Release (547dcf861cac0253a8abb52c135e44e02ba407a1, 2023-07-07) for Ubuntu Jammy), so I stick to the old spelling, but chances are that you will get a newer one to your computer that has switched the spelling to dash.
# above are knitr chunk options invisible in html
billionaires_df %>% ggplot() +
geom_smooth(mapping = aes(x = age, color = sex, y = daily_income)) +
facet_wrap(~ time)
7 You want axis ticks labeled denser
This is the default.
billionaires_df %>% ggplot() + geom_bar(aes(x = time, fill = sex))
You want ticks on the Y axis to mark each 500.
First, you need to know the right base R function to generate the sequence of the labels:
seq(from = 0, to = 3500, by = 500)[1] 0 500 1000 1500 2000 2500 3000 3500Then you tell ggplot2 that you want to modify the ticks on the Y-axis by calling the appropriate scale_…. function. Each aesthetic scale has its own scale_… function with arguments that modify its behavior. Note that the scale_ functions are divided into _discreet, _continuous, and sometimes _manual. You have to pick the type that corresponds to the type of the variable that you mapped on that aesthetic scale. The rule of the thumb is that categorical variables are always discrete and numeric variables are continuous. When you want a numeric variable to plot as discrete, you make it a factor before you pick the scale. In the plot, I also modify the X-axis.
billionaires_df %>% ggplot() +
geom_bar(aes(x = time, fill = sex)) +
scale_y_continuous(breaks = seq(from = 0, to = 3500, by = 500)) +
scale_x_continuous(breaks = seq(from = 2002, to = 2022, by = 2 ))
Years are conceptually a discrete variable, but they are numbers, so it is a continuous scale.
billionaires_df %>% ggplot() +
geom_bar(aes(x = time, fill = sex)) +
scale_x_continuous(breaks = seq(from = 2002, to = 2022, by = 1))
8 Axis tick labels overlap
When axis tick labels overlap, the easiest workaround is to tilt the plot by 90 degrees with coord_flip(). This function belongs to the coordinates layer in ggplot2.
billionaires_df %>% ggplot() +
geom_bar(aes(x = time, fill = sex)) +
scale_x_continuous(breaks = seq(from = 2002, to = 2022, by = 1)) +
coord_flip()
Sometimes this is all you have to do, but when you can’t do it, you must delve into the theme layer, calling the appropriate axis.text_… argument (for X, for Y, or for both). This wants that you call a function called element_text. This function has arguments with which you can tilt the labels, adjust their size and their position around the corresponding axis tick, etc.
What often helps is tilting the labels:
billionaires_df %>% ggplot() +
geom_bar(aes(x = time, fill = sex)) +
scale_x_continuous(breaks = seq(from = 2002, to = 2022, by = 1)) +
theme(axis.text.x = element_text(angle = 60))
Sometimes the labels bleed into the plot. This you can adjust with the vertical adjustment argument vjust. Experiment with the values. Normally you should need a small value like 0.1 or -0.1. Try vjust = 0 to get a reference point.
billionaires_df %>% ggplot() +
geom_bar(aes(x = time, fill = sex)) +
scale_x_continuous(breaks = seq(from = 2002, to = 2022, by = 1)) +
theme(axis.text.x = element_text(angle = 60, vjust = 0.7))
When you need to let the tick labels disappear, you do it like so:
billionaires_df %>% ggplot() +
geom_boxplot(aes(y = time)) +
scale_x_continuous(breaks = NULL) 
9 Change the titles of aesthetic scales, e.g. axes
Two ways to access the axis titles
scale_…xlab,ylab
9.1 scale_… to change the title of any aesthetic scale
billionaires_df %>% ggplot() +
geom_bar(aes(x = time, fill = sex)) +
scale_x_continuous(breaks = seq(from = 2002, to = 2022, by = 1),
name = "year") +
scale_y_continuous(name = "number of billionaires listed in Forbes\n and other popular charts") +
theme(axis.text.x = element_text(angle = 60, vjust = 0.7)) +
scale_fill_discrete(name = "gender", labels = c("women", "men", "diverse", "NA"))
9.2 xlab, ylab to change titles of the X, Y axes
These are shortcuts - probably written because these tasks are extremely common.
billionaires_df %>% ggplot() +
geom_bar(aes(x = time, fill = sex)) +
scale_x_continuous(breaks = seq(from = 2002, to = 2022, by = 1)) +
theme(axis.text.x = element_text(angle = 60, vjust = 0.7)) +
xlab(label = "year") +
ylab(label = "number of billionaires listed in Forbes\n and other popular charts")
# Notice the \n to enforce a new line10 Change things in color/fill with scale_…
Stick to ready-made options whenever you can, it can easily get too complicated.
Prefer to combine colors from color palettes designed by professionals. Some look awful but all guarantee enough contrast. Some of them even work for color-blind audience.
Important things to consider
- Discrete/ordinal or continuous variable? Some palettes work even only for binned variables (e.g. in `geom_hex`)
- Can values be distributed on a scale?
- diverging scale: two extremes (e.g., *totally agree - agree - don't know - disagree - totally disagree* or *-100 ....0 ....+100*)?
- sequential: zero to extreme (e.g., *beginner - intermediate - advanced - expert* or *0....100*)
- qualitative: no scale, just contrast (e.g., *bugs, beetles, butterflies*)10.1 scale_fill_brewer
for discrete variables
color palettes designed by cartographer Cynthia Brewer https://colorbrewer2.org/#type=sequential&scheme=BuGn&n=3
The palettes are divided into three types: diverging (
div), qualitative (qual) and sequential (seq)
# run repeatedly and explore the fill options with Help
billionaires_df %>% ggplot() +
geom_bar(aes(x = time, fill = sex)) +
scale_x_continuous(breaks = seq(from = 2002, to = 2022, by = 1)) +
theme(axis.text.x = element_text(angle = 60, vjust = 0.7)) +
scale_fill_brewer(name = "gender", type = "qual", palette = 8)
10.2 scale_fill_manual
your manually picked colors in a character vector
use either their names or HEX codes
easy to pick here https://r-charts.com/colors/
# run repeatedly and explore the fill options with Help
billionaires_df %>% ggplot() +
geom_bar(aes(x = time, fill = sex)) +
scale_x_continuous(breaks = seq(from = 2002, to = 2022, by = 1)) +
theme(axis.text.x = element_text(angle = 60, vjust = 0.7)) +
#scale_fill_discrete(type = c("red", "blue", "grey", "grey34"))
scale_fill_discrete(type = c("#A52A2A", "#4F94CD", "snow3", "#8B5A2B"))
11 To become a fill/color guru
read https://ggplot2-book.org/scales-colour#sec-colour-theory
ggplot2 functions implement color theory
which colors match/contrast
how they are rendered in computers
12 Add title to the entire plot
with
labswith
ggtitle
billionaires_df %>% ggplot() +
geom_bar(aes(x = time, fill = sex)) +
theme(axis.text.x = element_text(angle = 60, vjust = 0.7)) +
labs(title = "Count of world's billionaires over 2002-2022",
subtitle = "broken by sex")
billionaires_df %>% ggplot() +
geom_bar(aes(x = time, fill = sex)) +
scale_x_continuous(breaks = seq(from = 2002, to = 2022, by = 1)) +
theme(axis.text.x = element_text(angle = 60, vjust = 0.7)) +
ggtitle(label = "Count of world's billionaires over 2002-2022",
subtitle = "broken by sex")