Visualising uncertainty

Data

For this exercise, we’ll use data on coastal ocean temperatures by depth. The dataset was used as a TidyTuesday dataset so you can load it in directly from GitHub.

ocean_temperature <- readr::read_csv("https://raw.githubusercontent.com/rfordatascience/tidytuesday/main/data/2026/2026-03-31/ocean_temperature.csv")
ocean_temperature_deployments <- readr::read_csv("https://raw.githubusercontent.com/rfordatascience/tidytuesday/main/data/2026/2026-03-31/ocean_temperature_deployments.csv")

Download data

Alternatively, you can download the CSV files:

Download ocean_temperature CSV: ocean_temperature.csv
Download ocean_temperature_deployments CSV: ocean_temperature_deployments.csv

Load the data from your local copy:

ocean_temperature <- readr::read_csv("../data/ocean_temperature.csv")
ocean_temperature_deployments <- readr::read_csv("../data/ocean_temperature_deployments.csv")

For more information on the datasets, look at the data dictionaries.

Tip

You can copy the code to load the data on your laptop, and use your own R installation.

Or, you can go to nrennie.gitlab.io/dl4sg-uncertainty/code.html where you’ll find a live R environment with packages and data pre-installed.

Examples

Uncertainty over time with geom_ribbon()

plot_data <- readr::read_csv("../slides/data/linechart_3_data.csv")

# A tibble: 6 × 5
   Year Ethnicity                                  Estimate Upper Lower
  <dbl> <chr>                                         <dbl> <dbl> <dbl>
1  2012 Asian or Asian British                          4     6.8   1.1
2  2012 Black, African, Caribbean or Black British      5.6   9.3   1.8
3  2012 Mixed or Multiple ethnic groups                 1.6  10.5  -8.2
4  2012 Other ethnic group                              6.7  13.4  -0.7
5  2013 Asian or Asian British                          4.7   8.3   0.9
6  2013 Black, African, Caribbean or Black British      2.1   6.2  -2.3

library(ggplot2)
ggplot(
  data = plot_data,
) +
  geom_ribbon(
    mapping = aes(x = Year, ymin = Lower, ymax = Upper, fill = Ethnicity),
    alpha = 0.3
  ) +
  geom_line(
    data = plot_data,
    mapping = aes(x = Year, y = Estimate, colour = Ethnicity),
    linewidth = 1.5
  ) +
  scale_x_continuous(breaks = seq(2012, 2022, 2)) +
  labs(x = NULL) +
  theme_minimal()

Uncertainty in categorical estimates with geom_rect()

bar_data <- readr::read_csv("../slides/data/bar_chart_data.csv") |>
  dplyr::mutate(
    Category = factor(Category, levels = Category)
  ) |>
  dplyr::mutate(Category_num = as.numeric(factor(Category)))

# A tibble: 4 × 5
  Category             Estimate Lower Upper Category_num
  <fct>                   <dbl> <dbl> <dbl>        <dbl>
1 Degree or equivalent       16    14    18            1
2 Below degree level         12    10    14            2
3 Other qualification        10     5    15            3
4 No qualification           10     5    15            4

ggplot(data = bar_data) +
  geom_rect(
    mapping = aes(
      xmin = Lower, xmax = Upper,
      ymin = Category_num - 0.2, ymax = Category_num + 0.2
    ),
    fill = "deepskyblue4",
    alpha = 0.6
  ) +
  geom_point(
    mapping = aes(
      x = Estimate,
      y = Category_num
    ),
    pch = 23,
    fill = "white",
    colour = "deepskyblue4",
    size = 8
  ) +
  scale_y_continuous(
    breaks = unique(plot_data$Category_num),
    labels = levels(factor(plot_data$Category))
  ) +
  labs(y = NULL) +
  theme_minimal()

Warning: Unknown or uninitialised column: `Category_num`.

Warning: Unknown or uninitialised column: `Category`.

Exercise 1

Below is some R code to load and process some data:

library(ggplot2)
library(dplyr)
library(lubridate)
plot_data <- ocean_temperature |>
  filter(
    year(date) >= 2023, year(date) <= 2025,
    sensor_depth_at_low_tide_m %in% c(2, 5, 10)
  ) |>
  mutate(sensor_depth_factor = factor(sensor_depth_at_low_tide_m)) |>
  # compute CI
  mutate(
    lower = mean_temperature_degree_c - qnorm(0.975) * (sd_temperature_degree_c / sqrt(n_obs)),
    upper = mean_temperature_degree_c + qnorm(0.975) * (sd_temperature_degree_c / sqrt(n_obs))
  )

Then create a default line chart:

g <- ggplot(
  data = plot_data,
  mapping = aes(
    x = date, y = mean_temperature_degree_c,
    colour = sensor_depth_factor
  )
) +
  geom_line()
g

Inspect the data. You will see additional columns containing different levels of confidence intervals for different estimates in the data.
Decide if, and how, the uncertainty could be presented.
Edit the code to show uncertainty on the chart.
Hint: you may find geom_ribbon() in ggplot2 useful.

Exercise 2

Here are two charts. In groups, discuss:

What is good and bad about each of them?
How would you visualise this data?