Data Art & Science Blog

How to build a waffle chart with circle-shaped tiles using {waffle} and {ggplot2} libraries in R?

Juan Armando Torres Munguía — Sat, 14 Feb 2026 06:00:00 GMT

Overview

Waffle charts are a useful way to visualize part-to-whole relationships. Commonly, waffle charts depict a grid of regular squares to represent the distribution of a categorical variable.

Previously, in a post on this Blog¹, I built the following waffle chart to visualize disease outbreaks in the world between 2015 and 2024.

This visualization was developed in support of an analysis paper published in BMJ Global Health, which can be accessed here². As can be seen, this waffle chart uses the traditional approach of stacking colored squares to represent the number of outbreaks per year by pathogen. In this post, I will illustrate a visual alternative in which circles are used instead of squares to represent the distribution of diseases by year.

About the data

The information is sourced from the global dataset of pandemic- and epidemic-prone disease outbreaks³, whose data are freely available at the GitHub repository of the disease outbreaks project.

This global dataset documents more than 3,450 outbreaks across over 230 countries and territories from January 1996 to January 2026. The diseases are classified according to the International Classification of Diseases, 10th Revision (ICD-10), and the dataset contains information on the year, country, and pathogen for each outbreak.

The dataset of pandemic- and epidemic-prone disease outbreaks is also part of the Humanitarian Data Exchange coordinated by the United Nations Office for the Coordination of Humanitarian Affairs (OCHA).

Set-up

To create the waffle chart, we will use the following R packages:

# Data manipulation
library(tidyverse)

# Waffle chart
library(waffle)

# Visualization extensions for ggplot2
library(paletteer)
library(ggsci)
library(ggtext) 
library(shadowtext)
library(showtext)

# Create styled HTML/LaTeX tables
library(kableExtra)

Loading data

The table with the organized data can be downloaded here. This table contains the outbreaks recorded by year and by disease worldwide.

outbreaks_year_disease_grouped <- read.csv("outbreaks_year_disease_grouped.csv")

This table is shown below:

outbreaks_year_disease_grouped |>
  arrange(Year, -freq) |>
  kbl(caption = "Disease outbreaks per disease and year") |>
  kable_paper("hover", full_width = F)

Warning in attr(x, "align"): 'xfun::attr()' is deprecated.
Use 'xfun::attr2()' instead.
See help("Deprecated")

Disease outbreaks per disease and year
Year	icd104n	freq
2021	COVID-19	220
2021	Poliomyelitis	23
2021	Other	11
2021	Yellow fever	10
2021	Influenza	5
2021	Cholera	2
2021	Monkeypox	2
2021	Dengue	1
2022	COVID-19	230
2022	Monkeypox	53
2022	Other	39
2022	Hepatitis	38
2022	Cholera	29
2022	Yellow fever	12
2022	Poliomyelitis	6
2022	Dengue	5
2022	Influenza	5
2022	Measles	2
2023	COVID-19	210
2023	Dengue	60
2023	Other	40
2023	Cholera	19
2023	Yellow fever	12
2023	Influenza	9
2023	Measles	7
2023	Poliomyelitis	5
2023	Chikungunya	5
2023	Monkeypox	1
2024	COVID-19	142
2024	Dengue	95
2024	Other	32
2024	Yellow fever	13
2024	Influenza	10
2024	Monkeypox	8
2024	Poliomyelitis	1
2025	Influenza	121
2025	COVID-19	109
2025	Chikungunya	38
2025	Monkeypox	35
2025	Cholera	31
2025	Other	27
2025	Measles	7
2025	Yellow fever	5
2025	Poliomyelitis	2

Step 1. Visual elements of the plot

First, I define the font to be used in the final chart. To do this, I commonly use the font_add_google() function from the showtext package. package. This function retrieves font families from the Google Fonts repository. In this example, I use the Atkinson Hyperlegible Next font.

# Add custom font
font_add_google("Atkinson Hyperlegible Next", "Atkinson Hyperlegible Next") 
showtext_auto()

Then, I customize a theme to be applied to the plot.

# Custom theme for the waffle chart
theme_waffle_chart <- function() {

  # Introduce the previously selected font
  theme_minimal(base_family = "Atkinson Hyperlegible Next") +

  # Custom theme settings
  theme(
    # Axis settings
    axis.title = element_blank(), # Remove axis titles
    axis.line = element_blank(), # Remove axis lines
      
    # Title settings
    plot.title.position = "plot", # Position of the title
    plot.title = element_textbox(
      color = "black",
      face = "bold",
      size = 24,
      margin = margin(5, 0, 5, 0), # Top, right, bottom, left margins
      width = unit(1, "npc") # Full plot width
    ),

    plot.margin = unit(c(0.25, 0.25, 0.25, 0.25), "cm"),

    # Legend 
    legend.justification = c(1, 1),
    legend.title = element_text(face = "bold", size = 14),
    legend.title.position = "top",
    legend.text = element_text(face = "bold", size = 12),
    legend.direction = "vertical",
    legend.spacing.x = unit(30, "pt"),
    legend.key.size = unit(11, "pt"),
    legend.key.spacing.y = unit(3, "pt"),
    legend.key.spacing.x = unit(10, "pt"),

    # Subtitle settings
    plot.subtitle = element_textbox(
      color = "grey50",
      face = "bold",
      size = 18,
      margin = margin(0, 0, 40, 0), # Top, right, bottom, left margins
      width = unit(1, "npc")
    ),

    # Caption settings
    plot.caption = element_textbox(
      color = "grey70",
      size = 14,
      hjust = 0
    ),
    plot.caption.position = "plot",

    # Background and margins
    plot.background = element_rect(
      color = "white",
      fill = "white"
    ),
    panel.grid = element_blank(), 
    strip.text.x = element_text(face = "bold", margin = margin(t = 10), color = "black", size = 20),

    # Axis
    axis.ticks.y = element_line(linewidth = 1),
    axis.ticks.length.y = unit(5, "pt"),
    axis.text.x = element_text(face = "bold", color = "black", size = 12),
    axis.text.y = element_text(face = "bold", color = "black", size = 15),
    axis.title.x = element_text(face = "bold", margin = margin(t = 10), color = "black", size = 13),
    axis.title.y = element_text(face = "bold", margin = margin(r = 10), color = "black", size = 17)
    )
}

Then, I define the title, subtitle, and caption of the plot.

# Title, subtitle, and caption for the waffle chart
title_chart <- "Pandemic- and epidemic-prone disease outbreaks in the world | 2021–2025"
subtitle_chart <- "In 2025, influenza emerged as the most reported disease behind outbreaks, being responsible for 121 events."

For the caption, I use rich text by introducing markdown to format specific elements. This is enabled through the ggtext package.

caption_chart <- paste0(
  "**Note:** Each circle represents a country.",
  "
", 
  "**Data:** A global dataset of pandemic- and epidemic-prone disease outbreaks (DOI: 10.1038/s41597-022-01797-2).",
  "
", 
  "**Graphic:** Juan Torres Munguía."
  )

Step 2. Designing the waffle plot using `ggplot2` package

First, I use the geom_waffle() function to construct the waffle chart with squares. The main arguments of this function include size (border size of the tiles), n_rows (number of rows in the waffle grid), flip (orientation of the tiles), color (border color), and make_proportional (whether values are rescaled to proportions).

Additionally, to

ggplot(outbreaks_year_disease_grouped, 
      aes(fill = icd104n, values = freq)) +
  geom_waffle(size = 0.75, 
              n_rows = 10, 
              flip = TRUE,
              color = "white", 
              make_proportional = FALSE) +
  facet_wrap(~Year, 
             nrow = 1, 
             strip.position = "bottom")

In this example, I use the set of colors from the paletteMartin palette of the colorBlindness package. I also add the custom theme() along with the title, subtitle, and caption elements.

ggplot(outbreaks_year_disease_grouped, 
                aes(fill = icd104n, values = freq)) +
  geom_waffle(size = 0.75, 
              n_rows = 10, 
              flip = TRUE,
              color = "white", 
              make_proportional = FALSE) +
  facet_wrap(~Year, 
             nrow = 1, 
             strip.position = "bottom") +
  scale_fill_manual(values = c(paletteer_d("colorBlindness::paletteMartin"))) +
  scale_x_discrete() +
  scale_y_continuous(labels = function(x) x * 10, 
                     expand = c(0, 0)) +
  coord_equal() +
  labs(
    title = title_chart,
    subtitle = subtitle_chart,
    caption = caption_chart,
    x = "", 
    y = "Frequency of disease outbreaks",
    fill = "Pathogen:") +
  guides(fill = guide_legend(position = "right")) +
  theme_waffle_chart()

Step 3. Transforming the squares into circles.

Finally, to transform the tiles from squares into circles, I use radius as an aesthetic inside aes(), assigning the value grid::unit(0.5, “npc”). This produces circular tiles while preserving the waffle layout structure.

ggplot(outbreaks_year_disease_grouped, 
                aes(fill = icd104n, values = freq)) +
  geom_waffle(radius = grid::unit(0.5, "npc"),
              size = 0.75, 
              n_rows = 10, 
              flip = TRUE,
              color = "white", 
              make_proportional = FALSE) +
  facet_wrap(~Year, 
             nrow = 1, 
             strip.position = "bottom") +
  scale_fill_manual(values = c(paletteer_d("colorBlindness::paletteMartin"))) +
  scale_x_discrete() +
  scale_y_continuous(labels = function(x) x * 10, 
                     expand = c(0, 0)) +
  coord_equal() +
  labs(
    title = title_chart,
    subtitle = subtitle_chart,
    caption = caption_chart,
    x = "", 
    y = "Frequency of disease outbreaks",
    fill = "Pathogen:") +
  guides(fill = guide_legend(position = "right")) +
  theme_waffle_chart()

Step 4. Save the waffle chart as a high-quality image

showtext_opts(dpi = 320) # Resolution of 320 dpi for high-quality images ("retina")
ggsave(
  "waffle-pandemics-2026.png",
  dpi = 320,
  width = 14,
  height = 10,
  units = "in"
)
showtext_auto(FALSE)

References

Torres Munguía, J. A. A Waffle Chart of Disease Outbreaks in the World. (2025). doi:10.59350/kg492-0rv24.

Torres Munguía, J. A. & Martínez-Zarzoso, I. Global trends of pandemic-prone and epidemic-prone disease outbreaks in 2024. BMJ global health 11, e020708 (2026).

Torres Munguía, J. A., Badarau, F. C., Díaz Pavez, L. R., Martínez-Zarzoso, I. & Wacker, K. M. A global dataset of pandemic- and epidemic-prone disease outbreaks. Scientific data 9, 683 (2022).

Citation

BibTeX citation:

@online{torres_munguía2026,
  author = {Torres Munguía, Juan Armando},
  title = {How to Build a Waffle Chart with Circle-Shaped Tiles Using
    \{Waffle\} and \{Ggplot2\} Libraries in {R?}},
  date = {2026-02-14},
  url = {https://juan-torresmunguia.netlify.app/blog/posts/waffle-chart-disease-outbreaks-2025/},
  doi = {https://doi.org/10.59350/6ek9t-j8s96},
  langid = {en}
}

For attribution, please cite this work as:

Torres Munguía, J. A. How to build a waffle chart with circle-shaped tiles using {waffle} and {ggplot2} libraries in R? https://juan-torresmunguia.netlify.app/blog/posts/waffle-chart-disease-outbreaks-2025/ (2026) doi:https://doi.org/10.59350/6ek9t-j8s96.

Designing a map to visualize the geographic distribution of pandemic- and epidemic-prone disease outbreaks in 2025

Juan Armando Torres Munguía — Sun, 08 Feb 2026 06:00:00 GMT

Overview

Recently, I built a visualization as part of a paper published in BMJ Global Health, which can be accessedd here¹. The goal of this map was to communicate the geographic distribution of disease outbreaks worldwide during 2025. Additionally, I aimed to provide complementary information on the frequency of events at the continental level (suggested during the review process). In this blog post, I describe the process of reproducing this plot using {ggplot} and {sf} in R.

About the data

The statistical information is sourced from the global dataset of pandemic- and epidemic-prone disease outbreaks², whose data are freely available at the GitHub repository of the disease outbreaks project.

This global dataset documents over 3,450 outbreaks across more than 230 countries and territories from January 1996 to January 2026. The diseases are classified according to the International Classification of Diseases and the dataset contains information on the year, country, and pathogen for each outbreak.

Set-up

To create the map of the World, we will use the following R packages:

# Data manipulation
library(tidyverse)
library(janitor)

# Spatial data handling
library(sf) # Handle spatial vector data
library(rnaturalearth) # Access Natural Earth map data
library(rnaturalearthdata) # Support data for rnaturalearth

# Visualization
library(ggplot2)
library(ggspatial) # Spatial annotations for ggplot2
library(ggsflabel) # Labeling for sf objects
library(paletteer) # Access multiple color palettes
library(ggsci) # Scientific journal palettes
library(ggtext) # Rich text elements in ggplot
library(shadowtext) # Text with outline/shadow
library(showtext) # Custom fonts in plots
library(ggpubr) # Plot arrangement and publication utilities

# Tables
library(kableExtra) # Styled HTML/LaTeX tables

Loading data

The table with the organized data can be downloaded here. This table contains the outbreaks registered during 2025 by country and disease.

outbreaks_2025 <- read.csv("outbreaks_2025.csv")

The first five observations of the dataset are shown below:

outbreaks_2025 |>
 slice_head(n = 5) |>
 kbl(caption = "Disease outbreaks during 2025") |>
 kable_paper("hover", full_width = F)

Disease outbreaks during 2025
X	id_outbreak	Year	icd10n	icd103n	icd104n	icd10c	icd103c	icd104c	Disease	Definition	Country	iso2	iso3	unsd_region	unsd_subregion	who_region	DONs
1	2025AFGA000	2025	Intestinal infectious diseases	Cholera	Classical cholera	A00-A09	A00	A000	Cholera	Intestinal infection due to Vibrio cholerae	Afghanistan	AF	AFG	Asia	Southern Asia	Eastern Mediterranean Region	DON2928
2	2025AFGJ090	2025	Influenza and pneumonia	Influenza due to identified zoonotic or pandemic influenza virus	Influenza due to identified zoonotic or pandemic influenza virus	J09-J18	J09	J090	Influenza due to identified zoonotic or pandemic influenza virus	Influenza, caused by influenza virus strains of special epidemiological importance with an animal-human or inter-human transmission.	Afghanistan	AF	AFG	Asia	Southern Asia	Eastern Mediterranean Region	DON2938
3	2025AGOA000	2025	Intestinal infectious diseases	Cholera	Classical cholera	A00-A09	A00	A000	Cholera	Intestinal infection due to Vibrio cholerae	Angola	AO	AGO	Africa	Sub-Saharan Africa	African Region	DON2928, DON2912
4	2025ALBJ090	2025	Influenza and pneumonia	Influenza due to identified zoonotic or pandemic influenza virus	Influenza due to identified zoonotic or pandemic influenza virus	J09-J18	J09	J090	Influenza due to identified zoonotic or pandemic influenza virus	Influenza, caused by influenza virus strains of special epidemiological importance with an animal-human or inter-human transmission.	Albania	AL	ALB	Europe	Southern Europe	European Region	DON2938
5	2025ALBU071	2025	Provisional assignment of new diseases of uncertain etiology or emergency use	Emergency use of U07	COVID-19, virus identified	U00-U49	U07	U071	COVID-19	Infectious disease caused by the SARS-CoV-2 virus.	Albania	AL	ALB	Europe	Southern Europe	European Region	Coronavirus dashboard

Step 1. Data wrangling

First, it is necessary to reshape the outbreaks_2025 dataset to obtain the total number of outbreaks by country.

outbreaks_2025_country <- outbreaks_2025 |> 
 group_by(Country, iso3) |> 
 reframe(freq = n()) |> 
 ungroup() |> 
 arrange(-freq)

The first five observations of the table with outbreaks per country are:

outbreaks_2025_country |>
 slice_head(n = 5) |>
 kbl(caption = "Disease outbreaks by country during 2025") |>
 kable_paper("hover", full_width = F)

Disease outbreaks by country during 2025
Country	iso3	freq
China	CHN	7
Thailand	THA	6
Bangladesh	BGD	5
Bolivia (Plurinational State of)	BOL	5
France	FRA	5

I also compute the total number of outbreaks by continent.

outbreaks_2025_continent <- outbreaks_2025 |> 
 group_by(unsd_region) |>
 reframe(freq = n()) |>
 ungroup() |>
 arrange(-freq) |>
 rename(continent = unsd_region)

Here is the information at the continental level:

outbreaks_2025_continent |>
 kbl(caption = "Disease outbreaks by continent during 2025") |>
 kable_paper("hover", full_width = F)

Disease outbreaks by continent during 2025
continent	freq
Africa	103
Asia	87
Europe	87
Americas	73
Oceania	25

Step 2. Geographic information

To create the map with geographic boundaries, I use a shapefile located here. This file is imported with the st_read() function from the sf package.

shpsf <- st_read("shp/shp_outbreaks.shp", quiet = TRUE)
shpsf <- shpsf |> 
  filter(!is.na(continent))

# Disable s2 geometry processing to avoid projection issues
sf_use_s2(FALSE)

TThis shapefile can be directly plotted using the geom_sf() function of the sf package.

ggplot() +
 geom_sf(data = shpsf,
         color = "white",
         alpha = 1,
         linewidth = 0.1) +
  theme_bw()

However, this base map does not yet display the frequency of outbreaks by country. To incorporate that information, I merge the outbreaks_2025_country dataset (outbreaks per country) with the shapefile object shpsf.

outbreaks_shp_2025_country <- shpsf |>
 select(-Country) |>
 left_join(outbreaks_2025_country, by = "iso3") |>
 mutate(freq = replace_na(freq, 0)) |>
 select(-Country)

To promote accessibility, I apply a colorblind-friendly gradient scale to represent the minimum and maximum frequency of outbreaks.

ggplot() +
 
 geom_sf(
 data = outbreaks_shp_2025_country,
 aes(fill = freq),
 color = "white",
 alpha = 1,
 linewidth = 0.1
 ) +

 scale_fill_gradient(name = "Number of outbreaks", 
 low = "#FFFF6D", high = "#920000",
 breaks = seq(0, 10, by = 1)) +

 guides(
 fill = guide_legend(position = "top", nrow = 1)
 ) +
  theme_bw()

Identifying continents

Next, I define a set of colorblind-friendly colors to assign one color to each continent. A list of discrete palettes with accessible options can be found here.

continents <- shpsf |>
  select(continent) |>
  arrange(continent) |>
  st_drop_geometry() |>
  unique() 

continents_colors <- c("#920000", 
                       "#000000", 
                       "#004949", 
                       "#006DDB", 
                       "#490092")

continents <- continents |>
  cbind(continents_colors)

shpsf <- shpsf |>
 left_join(continents, by = c("continent"))

This produces a shapefile that includes the color assigned to each continent. I will use continental polygons to delimit only the perimeter of each continent. These objects can be obtained either through the ne_countries() function of the rnaturalearth package or by aggregating (e.g., dissolving) the original shpsf object.

# Map of Africa using ne_countries()
shpsf_africa <- ne_countries(continent = "Africa", returnclass = "sf") |>
 group_by(continent) |>
 summarise(geometry = st_union(geometry), 
 continent = unique(continent),
 continents_colors = "#920000") |>
 mutate(centroid = st_centroid(geometry))

# Map of the Americas using filter() and group_by()
shpsf_america <- shpsf |>
 filter(continent == "Americas") |>
 group_by(continent) |>
 summarise(geometry = st_union(geometry), 
 continent = unique(continent),
 continents_colors = unique(continents_colors)) |>
 mutate(centroid = st_centroid(geometry))

# Map of the the rest of continents using filter() and group_by()
shpsf_continent <- shpsf |>
 filter(continent != "Americas" & continent != "Africa") |>
 group_by(continent) |>
 summarise(geometry = st_union(geometry), 
 continent = unique(continent),
 continents_colors = unique(continents_colors)) |>
 mutate(centroid = st_centroid(geometry)) |>
 ungroup()

Given that the color information is stored in the continents_colors columns, I use scale_color_identity() so that the values in that column are interpreted directly as color codes. At this stage, the map looks as follows:

ggplot() +
 
 geom_sf(
 data = outbreaks_shp_2025_country,
 aes(fill = freq),
 color = "white",
 alpha = 1,
 linewidth = 0.1
 ) +

 scale_fill_gradient(name = "Number of outbreaks", 
 low = "#FFFF6D", high = "#920000",
 breaks = seq(0, 10, by = 1)) +

 guides(
 fill = guide_legend(position = "top", nrow = 1)
 ) +
 
 geom_sf(data = shpsf_continent,
 aes(color = continents_colors),
 fill = NA,
 size = 0.9,
 alpha = 1) +
 geom_sf(data = shpsf_america,
 aes(color = continents_colors),
 fill = NA,
 size = 0.9,
 alpha = 1) +
 geom_sf(data = shpsf_africa,
 aes(color = continents_colors),
 fill = NA,
 size = 0.9,
 alpha = 1) + 

 scale_color_identity() +

theme_bw()

During the review process, a suggestion was made to add text information to the map regarding the total number of outbreaks per continent. Therefore, I created an object containing the continental labels along with their longitude and latitude coordinates for positioning on the map.

continents_labels <- shpsf_continent |>
 rbind(shpsf_america, shpsf_africa) |>
 mutate(lon = st_coordinates(centroid)[, "X"], 
 lat = st_coordinates(centroid)[, "Y"]) |>
 st_drop_geometry() |>
 na.omit() |>
 left_join(outbreaks_2025_continent, by = "continent") |>
 rename("freq_2025" = freq)

Using annotate(), I add this text information to the map. In this case, the coordinates were manually adjusted to improve visual placement.

continents_labels <- continents_labels |>
 mutate(lat_y = c(30, 45, -45, 20, -10),
 lon_x = c(160, -30, 120, -130, -10))

ggplot() +
 
 geom_sf(
 data = outbreaks_shp_2025_country,
 aes(fill = freq),
 color = "white",
 alpha = 1,
 linewidth = 0.1
 ) +

 scale_fill_gradient(name = "Number of outbreaks", 
 low = "#FFFF6D", high = "#920000",
 breaks = seq(0, 10, by = 1)) +

 guides(
 fill = guide_legend(position = "top", nrow = 1)
 ) +
 
 geom_sf(data = shpsf_continent,
 aes(color = continents_colors),
 fill = NA,
 size = 0.9,
 alpha = 1) +
 geom_sf(data = shpsf_america,
 aes(color = continents_colors),
 fill = NA,
 size = 0.9,
 alpha = 1) +
 geom_sf(data = shpsf_africa,
 aes(color = continents_colors),
 fill = NA,
 size = 0.9,
 alpha = 1) +

 annotate(geom = "label", 
 fontface = "bold",
 linewidth = 1,
 fill = "white",
 color = continents_labels$continents_colors,
 x = continents_labels$lon_x, y = continents_labels$lat_y, 
 label = paste0(continents_labels$continent, " \n", continents_labels$freq_2025, " outbreaks"), 
 hjust = "center") + 

 scale_color_identity() +

theme_bw()

Step 3. Enhancing other visual elements of the plot

First, I define the font to be used in the final chart. I commonly use the font_add_google() function from the showtext package to import fonts from the Google Fonts repository. For the map in this post, I use the Roboto Condensed font.

# Add custom font
font_add_google("Roboto Condensed", "Roboto Condensed") 
showtext_auto()

Subsequently, I create a custom theme for the map.

# Custom theme for the map
theme_map_chart <- function() {

 # Introduce the previously selected font
 theme_minimal(base_family = "Roboto Condensed") +

 # Custom theme settings
 theme(
 # Axis settings
 axis.title = element_blank(), # Remove axis titles
 axis.line = element_blank(), # Remove axis lines
 axis.text = element_blank(), # Remove axis text
 
 # Title settings
 plot.title.position = "plot", # Position of the title
 plot.title = element_textbox(
 color = "black",
 face = "bold",
 size = 28,
 margin = margin(5, 0, 5, 0), # Top, right, bottom, left margins
 width = unit(1, "npc") # Full plot width
 ),

 plot.margin = unit(c(0, 0.25, 0.25, 0.25), "cm"),

 # Legend 
 legend.position.inside = c(0.5, 0.10),
 legend.title = element_text(face = "bold", size = 10),
 legend.title.position = "top",
 legend.text = element_text(face = "bold", size = 9),
 legend.text.position = "bottom",
 legend.direction = "horizontal",
 legend.spacing.x = unit(30, "pt"),
 legend.key.size = unit(20, "pt"),
 legend.key.spacing.y = unit(3, "pt"),
 legend.key.spacing.x = unit(0, "pt"),
 legend.background = element_rect(fill = "white", color = NA),
 
 # Subtitle settings
 plot.subtitle = element_textbox(
 color = "grey50",
 face = "bold",
 size = 12,
 margin = margin(0, 0, 5, 0), # Top, right, bottom, left margins
 width = unit(1, "npc")
 ),

 # Caption settings
 plot.caption = element_textbox(
 color = "grey70",
 size = 10,
 hjust = 0
 ),
 plot.caption.position = "plot",

 # Background and margins
 plot.background = element_rect(fill = "white", color = NA),

 panel.grid = element_blank(), 
 strip.text.x = element_text(face = "bold", margin = margin(t = 10), color = "black", size = 5)
 )
}

Then, I define the title, subtitle, and caption of the map.

# Title, subtitle, and caption for the map
title_chart <- "Geographic distribution of disease outbreaks in 2025"
subtitle_chart <- "At the continental level, Africa reported in 2025 the highest number of outbreaks with 103 events."

For the caption, I use rich text by introducing markdown to format specific elements.

caption_chart <- paste0(
 "**Data:** A global dataset of pandemic- and epidemic-prone disease outbreaks (DOI: 10.1038/s41597-022-01797-2).",
 "
", 
 "**Graphic:** Juan Torres Munguía."
 )

Step 4. Putting together all the previously designed elements

# Verificar el resultado
ggplot() +
 
 geom_sf(
 data = outbreaks_shp_2025_country,
 aes(fill = freq),
 color = "white",
 alpha = 1,
 linewidth = 0.1
 ) +

 scale_fill_gradient(name = "Number of outbreaks", 
 low = "#FFFF6D", high = "#920000",
 breaks = seq(0, 10, by = 1)) +

 guides(
 fill = guide_legend(position = "inside", nrow = 1)
 ) +
 
 geom_sf(data = shpsf_continent,
 aes(color = continents_colors),
 fill = NA,
 size = 0.9,
 alpha = 1) +
 geom_sf(data = shpsf_america,
 aes(color = continents_colors),
 fill = NA,
 size = 0.9,
 alpha = 1) +
 geom_sf(data = shpsf_africa,
 aes(color = continents_colors),
 fill = NA,
 size = 0.9,
 alpha = 1) +

 annotate(geom = "label", 
 fontface = "bold",
 linewidth = 1,
 fill = "white",
 color = continents_labels$continents_colors,
 x = continents_labels$lon_x, y = continents_labels$lat_y, 
 label = paste0(continents_labels$continent, " \n", continents_labels$freq_2025, " outbreaks"), 
 hjust = "center") + 

 scale_color_identity() +

labs(
    title = title_chart,
    subtitle = subtitle_chart,
    caption = caption_chart) +

 # Set the base theme without any background elements
 theme_map_chart()

I then save the file at a high-quality resolution:

showtext_opts(dpi = 320) # Resolution of 320 dpi for high-quality images ("retina")
ggsave(
  "outbreaks_map_colorblind_2026.png",
  dpi = 320,
  width = 30,
  height = 20,
  units = "cm"
)
showtext_auto(FALSE)

Final output is shown below:

References

Torres Munguía, J. A. & Martínez-Zarzoso, I. Global trends of pandemic-prone and epidemic-prone disease outbreaks in 2024. BMJ global health 11, e020708 (2026).

Torres Munguía, J. A., Badarau, F. C., Díaz Pavez, L. R., Martínez-Zarzoso, I. & Wacker, K. M. A global dataset of pandemic- and epidemic-prone disease outbreaks. Scientific data 9, 683 (2022).

Citation

BibTeX citation:

@online{torres_munguía2026,
  author = {Torres Munguía, Juan Armando},
  title = {Designing a Map to Visualize the Geographic Distribution of
    Pandemic- and Epidemic-Prone Disease Outbreaks in 2025},
  date = {2026-02-08},
  url = {https://juan-torresmunguia.netlify.app/blog/posts/world-map-outbreaks-2026/},
  langid = {en}
}

For attribution, please cite this work as:

Torres Munguía, J. A. Designing a map to visualize the geographic distribution of pandemic- and epidemic-prone disease outbreaks in 2025. https://juan-torresmunguia.netlify.app/blog/posts/world-map-outbreaks-2026/ (2026).

How has the number of persons seeking protection in Germany changed over time?

Juan Armando Torres Munguía — Fri, 01 Aug 2025 06:00:00 GMT

Overview

In Germany, as of 31 December 204, according to the Central Register of Foreigners, more than 14 million people of the 83.6 Million residents are foreigners.

It is important to note, that about 3.3 Million arrived to the country seeking protection. This figure is about 78% larger than the reported in 2020, and 343% more than 10 years ago.

An effective tool to analyze these data is through stacked area charts that allow to visualize the evolution of the variable of interest, persons seeking protection in this case, disaggregating it by categories. In this post, I will create a stacked area chart of this information using {ggplot2} in R.

About the data

Data are produced by the Central Register of Foreigners and are presented by Statistisches Bundesamt (Destatis) in their website in Table Persons seeking protection, by protection status, 2007 to 2024. I downloaded this information and saved it in an Excel file named persons-seeking-protection-germany.xlsx.

Set-up

First, I load the packages to be used:

library(tidyverse) 
library(kableExtra) 
library(readxl)
library(ggplot2) 
library(ggtext) 
library(ggrepel)
library(showtext) 
library(scales)

Loading data

I downloaded Table Persons seeking protection, by protection status, 2007 to 2024 and saved it in an Excel file named persons-seeking-protection-germany.xlsx.

seeking_protection <- read_excel("persons-seeking-protection-germany.xlsx")

The table looks like this:

seeking_protection |>
  kbl(
    caption = "Persons seeking protection, by protection status, 2007 to 2024"
    ) |>
  kable_paper("hover", full_width = F)

Persons seeking protection, by protection status, 2007 to 2024
Reference date	Population	Foreign population	Persons seeking protection	unsettled	recognised	temporary	permanent	refused
31 December 2007	82217837	6744880	457430	20145	301995	164350	137650	135290
31 December 2008	82002356	6727620	456050	18930	330365	150795	179570	106755
31 December 2009	81802257	6694775	477595	24620	361775	158735	203040	91195
31 December 2010	81751602	6753620	503470	35835	382325	168205	214115	85310
31 December 2011	80327900	6930895	505925	47130	373875	151045	222825	84920
31 December 2012	80523746	7213710	549825	65920	399050	165610	233440	84860
31 December 2013	80767463	7633630	613925	110335	410570	174110	236460	93020
31 December 2014	81197537	8152970	746320	177900	460140	208460	251675	108280
31 December 2015	82175684	9107895	1036235	349810	547935	285805	262130	138495
31 December 2016	82521653	10039080	1597570	574945	867500	599235	268265	155120
31 December 2017	82792351	10623940	1680700	348640	1154365	888355	266010	177700
31 December 2018	83019214	10915455	1781750	306095	1283225	1017760	265465	192430
31 December 2019	83166711	11228300	1839115	266470	1360070	1090475	269590	212575
31 December 2020	83155031	11432460	1856785	215960	1397685	1116970	280715	243140
31 December 2021	83237124	11817790	1936350	238945	1451375	1133545	317835	246030
31 December 2022	83118501	13383910	3078650	570060	2253875	1916630	337245	254710
31 December 2023	83456045	13895865	3173135	454795	2528935	2188665	340275	189405
31 December 2024	83577140	14061640	3304705	427415	2706320	2377115	329205	170970

Designing the stacked area chart

Before plotting, I reshape the data and rename some of their columns for easiness of working with them in the {ggplot2} context.

seeking_protection <- seeking_protection |>
  # I use year() to extract the information of the year
  mutate(year = year(dmy(`Reference date`))) |>
  # Only the subset of variables to use
  select(year, unsettled, temporary, permanent, refused) |>
  # Lengthening the data
  pivot_longer(!year, 
              names_to = "status", 
              values_to = "persons")

First, I set a theme for the chart and the font to “Josefin Sans” using the showtext package. More options can be found at https://fonts.google.com/.

font_add_google("Josefin Sans", "Josefin Sans")

showtext_auto()

theme_stacked_chart <- function() {
  theme_minimal(
    base_family = "Josefin Sans" 
  ) +
    # Custom settings
    theme(

      # Title settings
      plot.title.position = "plot", 
      plot.title = element_textbox(
        color = "#800000",
        face = "bold",
        size = 20,
        margin = margin(5, 0, 15, 0), # top, right, bottom, left
        hjust = 0,
        halign = 0      ),

      plot.subtitle = element_textbox(
      color = "#767676",
      #face = "bold",
      size = 18,
      margin = margin(5, 0, 35, 0),
      hjust = 0,
      halign = 0
      ),

      # Axis settings
      axis.line = element_line(
        size = 1.1, 
        colour = "#800000"
        ),

      axis.title.y = element_text(
        color = "#800000",
        face = "bold",
        size = 16
      ),
      axis.text.y = element_text(
        color = "#800000",
        face = "bold",
        size = 13
      ),
      axis.title.x = element_blank(),
      axis.text.x = element_text(
        color = "#800000",
        face = "bold",
        size = 13
      ),

      # Legend settings
      legend.position.inside = c(0.15, 0.80),
      legend.title = element_text(
        color = "#800000",
        face = "bold",
        size = 14
      ),
      legend.text = element_text(
        color = "#800000",
        size = 13
      ),

      # Caption settings
      plot.caption = element_text(
        color = "#800000",
        face = "italic",
        size = 14,
        hjust = 0,
        margin = margin(20, 0, 5, 0), # top, right, bottom, left
        ),

      plot.background = element_rect(
        color = "#D6D6CE",
        fill = "#D6D6CE"
      ),
      plot.margin = margin(40, 40, 40, 40) # top, right, bottom, left
    )
}

The plot is created with the following code:

seeking_protection <- seeking_protection |>
  mutate(status = str_to_title(status))

ggplot(seeking_protection,
  aes(
      x = year,
      y = persons,
      fill = status,
      color = status
      )
    ) +
  geom_area() +
  guides(fill = guide_legend(title = "Protection status", 
                             position = "inside"),
         color = guide_legend(title = "Protection status", 
                             position = "inside")
                             ) +
  scale_fill_manual(
    values = c("Unsettled"= "#767676",
               "Temporary"= "#CC8214", 
               "Refused" = "#8A9045",
               "Permanent" = "#155F83"
               )
  ) + 
  scale_color_manual(
    values = c("Unsettled"= "#767676",
               "Temporary"= "#CC8214", 
               "Refused" = "#8A9045",
               "Permanent" = "#155F83"
               )
  ) + 
  scale_y_continuous(
    labels = label_number(
      scale = 1e-6
      )
    ) +
  scale_x_continuous(
    expand = c(0, 0),
    limits = c(2006.5, 2024.5), 
    breaks = seq(2007, 2024, by = 1)
  ) +
  labs(
    title = "Persons seeking protection in Germany by status | 2007-2024",
    y = "Number of persons \n(in Millions)",
    caption = "Source: Central Register of Foreigners. \nDesign: Juan Torres Munguía."
    ) + 
  theme_stacked_chart()

showtext_opts(dpi = 320) 
ggsave(
  "stacked-area-germany-seeking-protection.png",
  dpi = 320,
  width = 12,
  height = 9,
  units = "in"
)
showtext_auto(FALSE) # Turn off the showtext functionality

Citation

BibTeX citation:

@online{torres munguía2025,
  author = {Torres Munguía, Juan Armando},
  title = {How Has the Number of Persons Seeking Protection in {Germany}
    Changed over Time?},
  date = {2025-08-01},
  url = {https://juan-torresmunguia.netlify.app/blog/posts/germany-persons-seeking-protection-stacked-area-chart},
  langid = {en}
}

For attribution, please cite this work as:

Torres Munguía, Juan Armando. 2025. “How Has the Number of Persons Seeking Protection in Germany Changed over Time?” August 1, 2025. https://juan-torresmunguia.netlify.app/blog/posts/germany-persons-seeking-protection-stacked-area-chart.

Creating a population pyramid to visualize crime victimization by age and sex in Mexico

Juan Armando Torres Munguía — Mon, 14 Jul 2025 06:00:00 GMT

Overview

Population pyramids are a powerful way to visualize demographic data, especially when analyzing age and sex patterns. In this post, I will elaborate a population pyramid using the {ggplot2} package in R, specifically focusing on crime victimization data from Mexico’s National Survey of Victimization and Perception of Public Safety (Encuesta Nacional de Victimización y Percepción sobre Seguridad Pública, ENVIPE).

Set-up

First, we need to install and load the necessary R packages.

library(tidyverse) # I always load tidyverse for data manipulation
library(ggplot2) # For data visualization
library(ggtext) # For formatting text in ggplot2
library(showtext) # For custom fonts in ggplot2
library(readxl) # For reading Excel files
library(kableExtra) # For data tables

Loading data

I will use the data from the table Population aged 18 and over by state, age group, sex and victimization condition (Población de 18 años y más por entidad federativa y grupos de edad según sexo y condición de victimización) from the 2024 ENVIPE available here.

envipe_data <- read.csv("victimization-age-sex-Mexico.csv")

Data looks like this:

envipe_data |>
  kbl(caption = "Prevalence of victimization by age and sex in Mexico, 2023") |>
  kable_paper("hover", full_width = F)

Prevalence of victimization by age and sex in Mexico, 2023
Age	Prevalence	Sex
18-19	25.7	Men
20-29	30.0	Men
30-39	27.3	Men
40-49	24.3	Men
50-59	20.9	Men
+60	15.1	Men
18-19	26.6	Women
20-29	30.5	Women
30-39	27.1	Women
40-49	22.6	Women
50-59	19.8	Women
+60	13.5	Women

Then, we use the {tidyverse} package to prepare the data for plotting.

envipe_data <- envipe_data |>
  mutate(Prevalence = case_when(Sex == "Women" ~ -1*Prevalence - 5,
                                TRUE ~ Prevalence + 5)) |>
  mutate(Age = factor(Age, levels = c("18-19", "20-29", "30-39", "40-49", "50-59", "+60")))

Creating the population pyramid

First, we estimate the adjusted limits for the x-axis.

prevalence_breaks <- seq(0, 30, by = 5)
# We add 5 to create a gap in the center of the plot
prevalence_breaks_adjusted <- c(prevalence_breaks + 5, -1*prevalence_breaks - 5)

Then, I create a custom theme for the chart and set the font to “Libre Franklin” using the showtext package. More font options are available https://fonts.google.com/.

font_add_google("Libre Franklin", "Libre Franklin")

showtext_auto()

# Custom theme for the chart
theme_pyramid_chart <- function() {
  theme_minimal(
    base_family = "Libre Franklin" 
  ) +
    # Custom theme settings
    theme(
      # remove grid lines
      panel.grid = element_blank(),

      # Axis settings
      axis.title.y = element_blank(),
      axis.text.y = element_blank(),
      axis.title.x = element_text(
        color = "white",
        face = "bold",
        size = 18
      ),
      axis.text.x = element_text(
        color = "white",
        face = "bold",
        size = 16
      ),

      # Title settings
      plot.title.position = "plot", 
      plot.title = element_textbox(
        color = "white",
        face = "bold",
        size = 24,
        margin = margin(5, 0, 5, 0), # top, right, bottom, left
        width = unit(1, "npc") 
      ),

      plot.subtitle = element_textbox(
      color = "white",
      face = "bold",
      size = 20,
      margin = margin(5, 0, 35, 0),
      width = unit(1, "npc")
    ),

      # Legend settings
      legend.position = "none",

      # Caption settings
      plot.caption = element_markdown(
        color = "white",
        face = "italic",
        size = 14,
        hjust = 0,
        margin = margin(50, 0, 5, 0) # top, right, bottom, left
        ),
      plot.background = element_rect(
        color = "#8C8D86",
        fill = "#8C8D86"
      ),
      plot.margin = margin(40, 40, 40, 40) # top, right, bottom, left
    )
}

title_chart <- "Crime victimization in Mexico | 2023"
subtitle_chart <- "Share of the population who were victims of crime by age group and sex"
caption_chart <- paste0("**Data:** 2024 National Survey of Victimization and Perception of Public Safety (ENVIPE)",
                        "
", 
                        "**Chart:** Juan Torres Munguía")

Finally, I use geom_col(), geom_label(), and annotate() from the {ggplot2} package to design the chart.

envipe_data |> 
  ggplot(aes(x = Age, 
             y = Prevalence, 
             fill = Sex)
        ) +
  geom_col(width = 1) +
  scale_fill_manual(
    values = c("Women" = "#E6C069", 
               "Men" = "#8DAB8E")) + 
  geom_label(
    aes(label = round(
      abs(Prevalence)-5, 1 # Round the figure 
      ), 
        y = Prevalence),
        color = "white",
        size = 5,
        fontface = "bold"
        ) +
  coord_flip(clip = "off") + 
  annotate(
    geom = "text",
    x = 6.75, 
    y = 7.5, 
    label = "Men",
    size = 8, 
    color = "#8DAB8E",
    fontface = "bold") +
  # Adding annotations for the sex of the victim
  annotate(
    geom = "text",
    x = 6.75, 
    y = -9.5, 
    label = "Women",
    size = 8, 
    color = "#E6C069",
    fontface = "bold") +
  # Adding a rectangle at the center of the plot
  # This rectangle will containg the x-axis labels (plot is inverted)
  # and is included between -5 and 5 values of the y-axis (plot is inverted)
  annotate(
    geom = "rect", 
    xmin = -Inf, 
    xmax = Inf, 
    ymin = 5, 
    ymax = -5,
    fill = "#8C8D86") + 
  # Labels for the vertical axis (age groups)
  annotate(
    geom = "text",
    x = c("18-19", "20-29", "30-39", "40-49", "50-59", "+60"), 
    y = 0, 
    label = c("18-19 \n years", "20-29", "30-39", "40-49", "50-59", "60 and \n older"),
    size = 6, 
    color = "white",
    fontface = "bold") +
  # I manually added a -40, 40 range of values to include the space in the center
  scale_y_continuous(
    limits = c(-40, 40),
    breaks = prevalence_breaks_adjusted,
    # Labels are renamed to be linked to real values of the x-axis, removing the 
    # space in the center
    labels = function(x) {abs(x) - 5}) + 
  labs(
    title = title_chart,
    subtitle = subtitle_chart,
    caption = caption_chart,
    x = "",
    y = "Proportion of victims (%)",
    fill = "") +
  theme_pyramid_chart()

# Set the resolution of the image 320 dpi is for high-quality images ("retina")
showtext_opts(dpi = 320) 
ggsave(
  "pyramid-crime-mexico.png",
  dpi = 320,
  width = 12,
  height = 9,
  units = "in"
)
showtext_auto(FALSE) # Turn off the showtext functionality

Citation

BibTeX citation:

@online{torres munguía2025,
  author = {Torres Munguía, Juan Armando},
  title = {Creating a Population Pyramid to Visualize Crime
    Victimization by Age and Sex in {Mexico}},
  date = {2025-07-14},
  url = {https://juan-torresmunguia.netlify.app/blog/posts/population-pyramid-mexico-crime},
  langid = {en}
}

For attribution, please cite this work as:

Torres Munguía, Juan Armando. 2025. “Creating a Population Pyramid to Visualize Crime Victimization by Age and Sex in Mexico.” July 14, 2025. https://juan-torresmunguia.netlify.app/blog/posts/population-pyramid-mexico-crime.

A dot plot of public acceptability of domestic violence in Cambodia between 2005 and 2009

Juan Armando Torres Munguía — Wed, 18 Jun 2025 06:00:00 GMT

Overview

According to official data from the National Institute of Statistics of Cambodia, in 2009 about 77 percent of Cambodians aged 15–49 considered it acceptable for a man to curse his wife. Although this figure is high, it represents a decrease of about 16 percentage points compared to the previous estimate from 2005. Analyzing these changes is important for gaining a better understanding of the phenomenon of violence against women and girls.

One of the most effective ways to visualize these changes over time is through dot plots. In this post, we will use the {ggplot2} package in R to create a dot plot that illustrates how public acceptance of domestic violence in Cambodia changed between 2005 and 2009.

Set-up

Load necessary libraries for data processing and visualization.

library(tidyverse) # For data wrangling
library(ggplot2) # Data visualization
library(readxl) # To read Excel files
library(kableExtra) # For data tables
library(ggplot2) # For data visualization
library(ggtext) # For formatting text in ggplot2
library(showtext) # For custom fonts in ggplot2
library(stringr) # For string manipulation

Loading data

Data comes from the report “Women and Men in Cambodia” produced by the National Institute of Statistics of the Ministry of Planning. This document provides insights into the situation of women and men in different spheres of life, including health, education, labour, decision making, and violence. The data used in this post is available in the table “Public Acceptability of Domestic Violence between 2005 and 2009” in section “VIOLENCE AGAINST WOMEN”. Report can be found here.

I extracted the data from the report and created an Excel file that can be found here

acceptability_violence <- read_excel("acceptability-violence-cambodia.xlsx")

This information looks like this:

acceptability_violence |>
  kbl(
    caption = "Public Acceptability of Domestic Violence in Cambodia between 2005 and 2009"
    ) |>
  kable_paper("hover", full_width = F)

Public Acceptability of Domestic Violence in Cambodia between 2005 and 2009
Action	2005	2009
Cursing	93	77
Pushing, shaking, or throwing something	74	46
Knocking on the head	69	36
Tying up or hitting	40	18
Threatening with a knife gun or another weapon	37	13
Burning or choking	24	10
Throwing acid	10	9

I use the pivot_longer() function in the {tidyverse} package to reshape the data to a long format, required for this plot.

acceptability_violence <- acceptability_violence |>
  pivot_longer(!Action, names_to = "Year", values_to = "Percentage")

Now the data looks like this:

acceptability_violence |>
  kbl(caption = "Public Acceptability of Domestic Violence in Cambodia between 2005 and 2009") |>
  kable_paper("hover", full_width = F)

Public Acceptability of Domestic Violence in Cambodia between 2005 and 2009
Action	Year	Percentage
Cursing	2005	93
Cursing	2009	77
Pushing, shaking, or throwing something	2005	74
Pushing, shaking, or throwing something	2009	46
Knocking on the head	2005	69
Knocking on the head	2009	36
Tying up or hitting	2005	40
Tying up or hitting	2009	18
Threatening with a knife gun or another weapon	2005	37
Threatening with a knife gun or another weapon	2009	13
Burning or choking	2005	24
Burning or choking	2009	10
Throwing acid	2005	10
Throwing acid	2009	9

Dot plot design

I create a custom theme for the dot plot and apply the font “Jacques Francois”. More font alternatives can be found here.

# Add custom font
font_add_google("Jacques Francois", "Jacques Francois") 

showtext_auto()

# Custom theme for the waffle chart
theme_dot_plot <- function() {
  theme_minimal(
    base_family = "Jacques Francois" # Base theme with custom font
  ) +
    # Custom theme settings
    theme(
      # remove grid lines
      panel.grid = element_line(
        size = 0.5,
        colour = "white"
        ),

      # Axis settings
      axis.line = element_blank(),
      
      axis.title.y = element_blank(),
      axis.text.y = element_text(
        color = "black",
        face = "bold",
        size = 16
      ),
      axis.title.x = element_text(
        color = "black",
        face = "bold",
        size = 16
      ),
      axis.text.x = element_text(
        color = "black",
        face = "bold",
        size = 16
      ),

      # Title settings
      plot.title.position = "plot", # Position of the title
      plot.title = element_text(
        color = "black",
        face = "bold",
        size = 24,
        margin = margin(5, 0, 5, 0) # Top, right, bottom, left
      ),
      
      # Subtitle settings
      plot.subtitle = element_text(
        color = "grey20",
        face = "italic",
        size = 20,
        margin = margin(0, 0, 10, 0)
        ),
      
      # Legend settings
      legend.position = "top",
      legend.title = element_blank(),
      legend.spacing.x = unit(0.2, "cm"),
      legend.key.spacing = unit(0.5, "cm"), # Spacing between legend keys
      legend.text = element_text(
        margin = margin(5, 2, 5, 0),
        face = "bold",
        color = "black",
        size = 16
      ),
      legend.direction = "horizontal",
      legend.byrow = FALSE,
      
      # Caption settings
      plot.caption = element_text(
        margin = margin(40, 0, 0, 0), # Top, right, bottom, left
        color = "black",
        size = 16
          ),
      plot.caption.position = "plot",

      plot.background = element_rect(
        color = "#FBB4AE",
        fill = "#FBB4AE"
      ),
      plot.margin = margin(20, 40, 20, 40)
    )
}

title_chart <- "Public acceptability of domestic violence in Cambodia"
subtitle_chart <- "Between 2005 and 2009, the share of people who tolerate violence has decreased"
caption_chart <- "Cambodia Key Gender Statistics, 2015 \nChart created by Juan Torres Munguía"

# Set the resolution of the image 320 dpi is for high-quality images ("retina")
showtext_opts(dpi = 320)

Finally, I create the dot plot using geom_line() and geom_point() functions from the {ggplto2} package.

dot_plot_acceptability <- acceptability_violence |> 
  ggplot(
    aes(x = Percentage, 
    y = Action)
    ) +
  geom_line(
    aes(group = Action),
    linewidth = 1.5,
    color = "#737373"
  ) +
  geom_point(
    aes(color = Year), 
    size = 6
  ) +
  scale_colour_manual(
    values = c("#084081", "#2B8CBE")
    ) +
  labs(
    title = title_chart, 
    subtitle = subtitle_chart,
    caption = caption_chart,
  ) +
  scale_x_continuous(
    limits = c(0, 100)
  ) +
  scale_y_discrete(
    # Wrap y-axis labels to a width of 32 characters
    labels = function(x) str_wrap(x, width = 32)
    ) +
  theme_dot_plot()

Finally, I export the plot using a 320-dpi resolution for high-quality (“retina”)

dot_plot_acceptability

showtext_opts(dpi = 320) 
ggsave(
  "cambodia-domestic-violence-dot-plot.png",
  dpi = 320,
  width = 12,
  height = 9,
  units = "in"
)

# Turn off the showtext functionality
showtext_auto(FALSE)

Citation

BibTeX citation:

@online{torres munguía2025,
  author = {Torres Munguía, Juan Armando},
  title = {A Dot Plot of Public Acceptability of Domestic Violence in
    {Cambodia} Between 2005 and 2009},
  date = {2025-06-18},
  url = {https://juan-torresmunguia.netlify.app/blog/posts/cambodia-acceptability-violence-dot-plot},
  langid = {en}
}

For attribution, please cite this work as:

Torres Munguía, Juan Armando. 2025. “A Dot Plot of Public Acceptability of Domestic Violence in Cambodia Between 2005 and 2009.” June 18, 2025. https://juan-torresmunguia.netlify.app/blog/posts/cambodia-acceptability-violence-dot-plot.

Three maps to examine the prevalence of intimate partner violence (IPV) in Cambodia

Juan Armando Torres Munguía — Sat, 10 May 2025 06:00:00 GMT

Overview

According to the 2021-22 Domestic Violence module of the Kingdom of Cambodia’s Demographic and Health Survey (CDHS), over 21% of Cambodian women aged 15–49 who have ever had a husband (or intimate partner), have experienced physical, sexual, economic, and/or emotional intimate partner violence in their lifetime. This figures varies significantly across the country, with some provinces reporting rates as high as 45% while others report rates as low as 6%.

A better understanding of the geographical distribution of intimate partner violence (IPV) in Cambodia can help policymakers and organizations to target their efforts more effectively. In this blog post, I will show how to create choropleth maps using R and shapefiles to visualize the prevalence of IPV in Cambodia at the provincial level. Additionally, I will use the {patchwork} package to combine these maps into the same graphic. The result will look like this:

About the data

Information comes from the document “Further Data Analysis Report. From the Cambodia Demographic and Health Survey 2021-2022”, available here. In particular, I will use Table A.1: Intimate-partner violence by background characteristics.

Set-up

First, we install and load the necessary R packages.

library(tidyverse) 
library(sf) # For handling shapefiles
library(readxl) # To read Excel files
library(kableExtra) # For data tables
library(ggplot2) 
library(ggtext) 
library(showtext) 
library(patchwork) # To combine multiple ggplots
library(colorspace) # For color palettes
library(grid) # To add lines and text elements to graphical objects

Loading data

I saved the information from Table A.1: Intimate-partner violence by background characteristics in the Excel file intimate-partner-violence-cambodia.xlsx.

ipv_cambodia <- read_excel("intimate-partner-violence-cambodia.xlsx")

This information looks like this:

ipv_cambodia |>
  kbl(
    caption = "Lifetime experience of specific acts of intimate partner physical violence committed by their current or most recent partner by provinces, Cambodia 2021-22"
    ) |>
  kable_paper("hover", full_width = F)

Lifetime experience of specific acts of intimate partner physical violence committed by their current or most recent partner by provinces, Cambodia 2021-22
Region	Emotional violence	Physical violence	Sexual violence	Physical or sexual	Physical or sexual or emotional	Number of women who ever had a husband/intimate partner
Banteay Meanchey	31.3	18.2	4.6	20.2	34.9	235
Battambang	24.9	16.1	6.3	19.8	30.4	378
Kampong Cham	7.7	5.2	1.1	5.8	11.5	352
Kampong Chhnang	19.2	6.7	1.7	6.7	19.2	213
Kampong Speu	16.6	9.5	2.6	10.1	19.1	362
Kampong Thom	32.4	12.1	5.0	14.4	35.1	242
Kampot	20.8	6.4	4.0	9.0	23.0	218
Kandal	23.9	6.2	2.0	6.5	23.9	423
Koh Kong	5.4	3.8	0.4	3.8	6.2	41
Kratie	13.7	9.1	2.8	10.2	15.9	138
Mondul Kiri	42.7	19.0	3.6	19.4	45.2	34
Phnom Penh	10.3	5.1	1.3	5.8	11.0	904
Preah Vihear	38.1	20.2	8.5	22.4	40.0	108
Prey Veng	25.0	7.2	1.7	8.6	26.3	366
Pursat	10.8	3.6	0.7	3.6	10.8	111
Ratanak Kiri	24.7	6.6	0.8	7.1	25.7	96
Siemreap	21.5	12.0	2.9	13.5	24.6	482
Preah Sihanouk	10.3	3.4	1.6	4.0	10.8	75
Stung Treng	29.3	12.5	7.5	16.3	31.5	66
Svay Rieng	10.9	8.3	2.1	8.8	13.6	217
Takeo	18.4	7.4	1.2	8.0	19.3	344
Otdar Meanchey	15.1	6.1	0.0	6.1	15.4	77
Kep	8.0	7.2	1.2	7.7	11.0	18
Pailin	11.3	9.5	1.7	9.5	14.2	29
Tboung Khmum	11.1	7.3	2.3	7.3	14.0	252

Now, we load geographical data of the administrative division of Cambodia. These data are obtained from the Humanitarian Data Exchange (HDX) initiative of the United Nations Office for the Coordination of Humanitarian Affairs (OCHA). The information comes as a shapefile .shp and is available here in a zip file named som_adm_ocha_20250108_AB_SHP.zip. I downloaded this file and located it in /khmadm_gov_20181004_ab_shp folder.

shp_cambodia <- read_sf("khmadm_gov_20181004_ab_shp/khm_admbnda_adm1_gov_20181004.shp")

Data wrangling

Next step is to combine the geographic information from the shapefile, shp_cambodia, with the data frame with the information on IPV, ipv_cambodia. To achieve this goal, we need to make sure that both datasets have a common variable to merge them. In this case, we will use the province name ADM1_EN in the shapefile and Region in the data frame. Then, I compare if the names of the provinces in both datasets match:

setdiff(shp_cambodia$ADM1_EN, ipv_cambodia$Region)

[1] "Oddar Meanchey"

The output indicates that there is one discrepancy. In shp_cambodia, the name of the province “Oddar Meanchey” corresponds to “Otdar Meanchey” in ipv_cambodia. I will rename this province in the shapefile to match the name in the data from the official report.

shp_cambodia <- shp_cambodia |>
  mutate(ADM1_EN = case_when(
    ADM1_EN == "Oddar Meanchey" ~ "Otdar Meanchey",
    TRUE ~ ADM1_EN
  ))

After this modification, I join both datasets using a left join to keep all the provinces in the shapefile and add the corresponding IPV values from the data frame:

shp_cambodia <- shp_cambodia |>
  left_join(ipv_cambodia, by = c("ADM1_EN" = "Region"))

Creating the map

My idea is to develop three maps, each representing one type of violence against women and girls: emotional, physical, and sexual. I will first create the map corresponding to emotional violence.

# More font alternatives can be found here https://fonts.google.com/
font_add_google("Scada", "Scada") 

showtext_auto()

map_ipv_emotional_cambodia <- ggplot() +
  geom_sf(
    data = shp_cambodia, 
    aes(fill = `Emotional violence`), 
    color = "white", # Color of the province borders
    linewidth = 0.5
  ) +
  geom_sf_text(
    data = shp_cambodia,
    aes(label = `Emotional violence`),
    check_overlap = TRUE, # To avoid overlaping text in the plot
    size = 4, 
    color = "black",
    fontface = "bold"
  ) +
  scale_fill_gradient2(
    low = "#FFFE9EB3",
    mid = "#C53270B3",
    high = "#040404B3",
    midpoint = 22.5,
    limits = c(0, 45),
    breaks = c(0, 15, 30, 45),
    labels = c("0", "15", "30", "45")
  ) +
  labs(
    title = "Emotional", # Title of this subplot
    subtitle = "",
    caption = "",
    x = "",
    y = "",
    fill = "Percentage of victims (%)" # Title of the legend
  ) +
  theme_minimal(
    base_family = "Scada"
  ) +
  theme(
    plot.margin = margin(0, 40, 0, 40),
    
    plot.background = element_blank(), # No color in the background
    plot.title = element_text(
      color = "black",
      face = "bold",
      size = 22,
      hjust = 0.5,
      margin = margin(5, 0, 5, 0)
    )
  )

Now, I replicate the same format for physical and sexual violence.

map_ipv_physical_cambodia <- ggplot() +
  geom_sf(
    data = shp_cambodia, 
    aes(fill = `Physical violence`), 
    color = "white", 
    linewidth = 0.5
  ) +
  geom_sf_text(
    data = shp_cambodia,
    aes(label = `Physical violence`),
    check_overlap = TRUE,
    size = 4, 
    color = "black",
    fontface = "bold"
  ) +
  scale_fill_gradient2(
    low = "#FFFE9EB3",
    mid = "#C53270B3",
    high = "#040404B3",
    midpoint = 22.5,
    limits = c(0, 45)
  ) +
  labs(
    title = "Physical",
    subtitle = "",
    caption = "",
    x = "",
    y = "",
    fill = ""
  ) + 
  guides(fill = "none") +
  theme_minimal(
    base_family = "Scada"
  ) +
  theme(
    plot.margin = margin(0, 40, 0, 40),
    
    plot.background = element_blank(),
    plot.title = element_text(
      color = "black",
      face = "bold",
      size = 22,
      hjust = 0.5,
      margin = margin(5, 0, 5, 0)
    )
  )

map_ipv_sexual_cambodia <- ggplot() +
  geom_sf(
    data = shp_cambodia, 
    aes(fill = `Sexual violence`), 
    color = "white", 
    linewidth = 0.5
  ) +
  geom_sf_text(
    data = shp_cambodia,
    aes(label = `Sexual violence`),
    check_overlap = TRUE,
    size = 4, 
    color = "black",
    fontface = "bold"
  ) +
  scale_fill_gradient2(
    low = "#FFFE9EB3",
    mid = "#C53270B3",
    high = "#040404B3",
    midpoint = 22.5,
    limits = c(0, 45)
  ) +
  labs(
    title = "Sexual",
    subtitle = "",
    caption = "",
    x = "",
    y = "",
    fill = ""
  ) + 
  guides(fill = "none") +
  theme_minimal(
    base_family = "Scada"
  ) +
  theme(
    plot.margin = margin(0, 40, 0, 40),
    
    plot.background = element_blank(),
    plot.title = element_text(
      color = "black",
      face = "bold",
      size = 22,
      hjust = 0.5,
      margin = margin(5, 0, 5, 0)
    )
  )

Now, using {patchwork} package, I combine the three plots into one single representation. I also use plot_annotation() function to add a title, subtitle, and caption to the overall design.

# Design display two rows, one with the three maps, and the second one only with the common guide
map_ipv_cambodia <- (map_ipv_emotional_cambodia + map_ipv_physical_cambodia + map_ipv_sexual_cambodia) / guide_area() + 
  # Custom title, subtitle, and caption
  plot_annotation(
    title = "Intimate partner violence in Cambodia, 2021-22",
    subtitle = "Proportion of ever-partnered women and girls aged 15-49 who have been victims of violence perpetrated by their intimate partner, by type of violence.",
    caption = "**Data:** 2021-22 Cambodia Demographic and Health Survey (CDHS) 
 **Chart:** Juan Torres Munguía",
    theme = theme(
      # Title settings
      plot.title.position = "plot", 
      plot.title = element_text(
        color = "black",
        face = "bold",
        size = 30,
        hjust = 0,
        margin = margin(5, 0, 5, 50)
      ),
      
      plot.subtitle = element_text(
        color = "grey20",
        size = 25,
        hjust = 0,
        margin = margin(5, 0, 40, 50)
        ),
        
        # Caption settings
        # I use element_markdown() because the caption has markdown text
        plot.caption = element_markdown(
          color = "black",
          size = 15,
          hjust = 0,
          margin = margin(80, 0, 5, 50)
        )
  )) +

  plot_layout(guides = "collect",
              heights = unit(c(13, 1),
               c('cm', 'cm'))
              ) + 
  theme_minimal(
    base_family = "Scada"
  ) &
  # Custom theme settings
  theme(
    legend.background = element_blank(),
    legend.position = "bottom",
    legend.title = element_text(
      margin = margin(5, 0, 10, 0),
      face = "bold",
      color = "grey10",
      size = 14,
      hjust = 0.5
    ),
    legend.title.position = "top",
    legend.text = element_text(
      margin = margin(5, 5, 5, 5), 
      face = "bold",
      color = "grey10",
      size = 14
    ),
    legend.key.width = unit(1.4, "cm"),
    legend.key.height = unit(1, "cm"),
    legend.direction = "horizontal",

    plot.background = element_rect(
      color = "#E5E5E5",
      fill = "#E5E5E5"),
    # Axis settings
    axis.title = element_blank(),
    axis.line = element_blank(),
    axis.text = element_blank(),
    axis.ticks = element_blank(),
    panel.grid = element_blank()
    
  )

Finally, I use grid.lines() from {grid} package to add a horizontal line after the subtitle.

map_ipv_cambodia

grid.lines(
  x = c(0.04, 0.96), # Values should be between 0 and 1
  y = 0.88, # Value should be between 0 and 1
  gp = gpar(col = "#C53270B3", 
            lwd = 4)
)

Citation

BibTeX citation:

@online{torres munguía2025,
  author = {Torres Munguía, Juan Armando},
  title = {Three Maps to Examine the Prevalence of Intimate Partner
    Violence {(IPV)} in {Cambodia}},
  date = {2025-05-10},
  url = {https://juan-torresmunguia.netlify.app/blog/posts/map-intimate-partner-violence-cambodia},
  langid = {en}
}

For attribution, please cite this work as:

Torres Munguía, Juan Armando. 2025. “Three Maps to Examine the Prevalence of Intimate Partner Violence (IPV) in Cambodia.” May 10, 2025. https://juan-torresmunguia.netlify.app/blog/posts/map-intimate-partner-violence-cambodia.

Working with Humanitarian Exchange Language (HXL) Hashtags

Juan Armando Torres Munguía — Wed, 26 Mar 2025 06:00:00 GMT

Setting Up the Environment

We begin by loading the necessary R packages:

# Load required packages
library(jsonlite) # For working with JSON data
library(httr) # For HTTP requests
library(dplyr) # For data manipulation
library(readxl) # For reading Excel files
library(writexl) # For writing Excel files

Step 1: Accessing the Dataset

We retrieve the latest dataset of disease outbreaks from GitHub:

# Define the API URL from GitHub containing the latest dataset update
url_api <- "https://api.github.com/repos/jatorresmunguia/disease_outbreak_news/contents/Last%20update"

# Retrieve the list of files in the "Last update" folder
file_list <- fromJSON(content(GET(url_api), as = "text"))$name

# Filter the most recent file that starts with "outbreaks"
last_file <- file_list[grepl(file_list, pattern = paste0("^outbreaks"))]

# Select the file with a .csv extension
rdata_file <- last_file[grepl(".csv$", last_file)]

# Construct the CSV file URL from GitHub and read the data into a dataframe
last_version <- read.csv(paste0("https://raw.githubusercontent.com/jatorresmunguia/disease_outbreak_news/refs/heads/main/Last%20update/", rdata_file), header = TRUE)

# Remove unnecessary ICD-11 classification columns
last_version <- last_version |> select(-c(icd11c1, icd11c2, icd11c3, icd11l1, icd11l2, icd11l3))

Step 2: Adding WHO and UNSD Regional Classifications

We integrate additional geographic classifications from WHO and UNSD.

WHO Regions

# Define the URL for the WHO regional classification dataset
url_who_class <- "https://cdn.who.int/media/docs/default-source/air-pollution-documents/air-quality-and-health/un-agencies-region-classification-for-country.xlsx?sfvrsn=289af35f_3"

# Create a temporary file to store the downloaded Excel file
temp_file <- tempfile(fileext = ".xlsx")

# Download the WHO classification file and save it to the temporary file
GET(url_who_class, write_disk(temp_file, overwrite = TRUE))

Response [https://cdn.who.int/media/docs/default-source/air-pollution-documents/air-quality-and-health/un-agencies-region-classification-for-country.xlsx?sfvrsn=289af35f_3]
  Date: 2025-04-28 06:21
  Status: 200
  Content-Type: application/vnd.openxmlformats-officedocument.spreadsheetml.sheet
  Size: 28.9 kB
  C:\Users\jator\AppData\Local\Temp\RtmpiGD1DX\file990423923f25.xlsx

# Read the WHO classification data from the downloaded Excel file
who_class <- read_excel(temp_file)

# Select relevant columns (ISO country code and WHO region name)
# Rename columns for consistency with the main dataset
last_version <- who_class |>
  select(`ISO Country code`, `WHO Region name2`) |>
  rename(iso3 = `ISO Country code`, who_region = `WHO Region name2`) |>
  right_join(last_version, by = "iso3") # Merge with the main dataset based on the ISO3 country code

UNSD Regions

# Read the UNSD regional classification dataset from an Excel file
# The dataset is available at: https://unstats.un.org/unsd/methodology/m49/
unsd_class <- read_excel(path = "UNSD — Methodology.xlsx")

# Select relevant columns: ISO-alpha3 country code, region name, and sub-region name
# Rename columns to match the main dataset
last_version <- unsd_class |>
  select(`ISO-alpha3 Code`, `Region Name`, `Sub-region Name`) |>
  rename(
    iso3 = `ISO-alpha3 Code`,
    unsd_region = `Region Name`,
    unsd_subregion = `Sub-region Name`
  ) |>
  right_join(last_version, by = "iso3") # Merge with the main dataset based on the ISO3 country code

Step 3: Structuring the Dataset

To facilitate better organization, we:

Order the dataset by year, country, and disease code.
Assign a unique outbreak ID.
Insert an empty row for HXL headers.

# Arrange the dataset in descending order by Year, then by ISO3 country code, and ICD-10 4-character code
last_version <- last_version |>
  arrange(desc(Year), iso3, icd104c) |>
  # Create a unique outbreak ID by concatenating Year, ISO3 country code, and ICD-10 4-character code
  mutate(id_outbreak = paste0(Year, iso3, icd104c)) |>
  # Add an empty row at the beginning of the dataset (for later HXL hashtag insertion)
  add_row(.before = 1)

Step 4: Adding HXL Hashtags

We define HXL hashtags for each relevant column:

# Define the mapping of column names to their corresponding HXL hashtags
cols2hxl <- c(
  id_outbreak = "#Year+iso3+icd4", # Unique outbreak ID
  Country = "#country+name", # Country name
  iso2 = "#country+code+iso2", # ISO 2-letter country code
  iso3 = "#country+code+iso3", # ISO 3-letter country code
  unsd_region = "#region+unsd", # UNSD region
  unsd_subregion = "#subregion+unsd", # UNSD subregion
  who_region = "#region+who", # WHO region
  Year = "#date+year", # Year of the outbreak
  icd10n = "#disease+name+icd", # Disease name (ICD-10)
  icd103n = "#disease+name+icd3", # Disease name (ICD-10 3-character)
  icd104n = "#disease+name+icd4", # Disease name (ICD-10 4-character)
  icd10c = "#disease+code+icd", # Disease code (ICD-10)
  icd103c = "#disease+code+icd3", # Disease code (ICD-10 3-character)
  icd104c = "#disease+code+icd4", # Disease code (ICD-10 4-character)
  Disease = "#disease+name", # Disease name
  DONs = "#news+id", # News ID
  Definition = "#x_disease+definition" # Disease definition
)

# Add HXL hashtags to the relevant columns in the dataset
last_version <- last_version |>
  mutate(across(all_of(names(cols2hxl)), ~ replace(., 1, cols2hxl[cur_column()]))) |>
  # Select the relevant columns for the final dataset
  select(id_outbreak, Year, icd10n, icd103n, icd104n, icd10c, icd103c, icd104c, Disease, Definition, Country, iso2, iso3, unsd_region, unsd_subregion, who_region, DONs)

Step 5: Exporting the Final Dataset

Finally, we save the dataset as an Excel file:

write_xlsx(x = list(Data = last_version), path = "disease_outbreaks_HDX.xlsx")

Citation

BibTeX citation:

@online{torres munguía2025,
  author = {Torres Munguía, Juan Armando},
  title = {Working with {Humanitarian} {Exchange} {Language} {(HXL)}
    {Hashtags}},
  date = {2025-03-26},
  url = {https://juan-torresmunguia.netlify.app/blog/posts/hdx-tags-to-outbreaks-data},
  langid = {en}
}

For attribution, please cite this work as:

Torres Munguía, Juan Armando. 2025. “Working with Humanitarian Exchange Language (HXL) Hashtags.” March 26, 2025. https://juan-torresmunguia.netlify.app/blog/posts/hdx-tags-to-outbreaks-data.

Visualizing armed conflict using ACLED’s data and HDX shapefiles

Juan Armando Torres Munguía — Sun, 09 Mar 2025 06:00:00 GMT

Overview

ACLED (Armed Conflict Location & Event Data Project) provides detailed data on political violence and protest events across the globe. This blog post explores the visualization of armed conflict in the Lake Chad Basin, focusing on the dynamics of violence against civilians in the region. Using ACLED’s data in combination with shapefiles from the Humanitarian Data Exchange (HDX) initiative of the United Nations Office for the Coordination of Humanitarian Affairs (OCHA), we leverage {ggplot2}, {sf} and the {tidyverse} to create insightful maps that highlight the impact of armed conflict.

Set-up

Load necessary libraries for data processing and visualization.

library(tidyverse) # For data manipulation and visualization
library(sf) # For handling spatial data
library(ggplot2) # For data visualization
library(patchwork) # For combining plots
library(ggtext) # Use rich text elements in ggplot
library(showtext) # Add custom fonts to plots
library(ggspatial) # For interacting with spatial data using ggplot2
library(rnaturalearth) # To provide map data that can be visualized using other R packages
library(rnaturalearthdata) # To store data for the rnaturalearth package

# Set environment to English (useful for date/time formatting)
invisible(Sys.setenv(LANG = "en"))
invisible(Sys.setlocale("LC_TIME", "English"))

Loading data

Accessing ACLED data requires registering on their ACLED Access Portal to generate a unique access key. Once registered, you can download the data using their Data Export Tool. I downloaded all the events occurred in Africa since 1997 and save them in file acled_africa_1997-2025.csv. To have a better understanding of their data, they provide a codebook.

conflict_data <- read.csv("acled_africa_1997-2025.csv")

Information of the national and sub national administrative boundaries in the Lake Chad Basin is obtained from the HDX. The information comes as a shapefile .shp and is available here in a zip file named lcb_admbnda_adm2_ocha.zip. The shapefile contains the administrative boundaries of Lake Chad Basin region at the subnational level. In particular, we will use file lcb_admbnda_adm2_ocha.shp. I also downloaded the administrative boundaries of the West and Central Africa region available here.

# Lake Chad Basin Area - Administrative boundaries for Nigeria, Chad, Cameroon and Niger
shp_lake_chad_basin <- read_sf("lcb_admbnda_adm2_ocha/lcb_admbnda_adm2_ocha.shp")

# West and Central Africa - Administrative boundaries levels 0 - 2 and Settlements
west_central_africa <- read_sf("wca_admbnda_adm2_ocha/wca_admbnda_adm2_ocha.shp")

Data processing

As indicated in the ACLED’s codebook, geographic location of the events is specified in columns latitude and longitude in the conflict_data object and they are using the EPSG:4326 coordinate reference system. We have to ensure, that both shp_lake_chad_basin and west_central_africa have the same coordinate reference system. To achieve this, we use st_transform() from {sf} package.

# st_transform to convert coordinates to ACLED's data coordinate reference system
shp_lake_chad_basin <- st_transform(shp_lake_chad_basin,
  crs = 4326
)
west_central_africa <- st_transform(west_central_africa,
  crs = 4326
)

Now, we subset the events to exclusively account for those occurred between 2020 and 2024 in the Lake Chad Basin. To achieve this, we use the {tidyverse} package for data wrangling and {sf} for spatial filtering. By applying the st_filter function, we retain only events located within the Lake Chad Basin area.

conflict_data <- conflict_data |>
  # Keep only events from 2020 to 2024
  filter(year >= 2020 & year < 2025) |>
  # Select only "Violence against civilians" events of type "Attack"
  filter(event_type == "Violence against civilians" & sub_event_type == "Attack") |>
  # Keep events from Nigeria, Niger, Cameroon, and Chad
  filter(country %in% c("Nigeria", "Niger", "Cameroon", "Chad")) |>
  # Extract the month name from event_date
  mutate(
    month = month(dmy(event_date),
      label = TRUE
    )
  ) |>
  # Select only relevant columns
  select(
    event_id_cnty, event_date, month, year, event_type, sub_event_type,
    actor1, actor2, location, latitude, longitude, fatalities, notes
  )

# Convert the data to a spatial object using latitude and longitude (WGS 84 CRS)
conflict_data <- st_as_sf(conflict_data,
  coords = c("longitude", "latitude"),
  crs = 4326
)

# Filter events that fall within the Lake Chad Basin shapefile
conflict_data_lake_chad <- conflict_data |>
  st_filter(shp_lake_chad_basin)

We then processes administrative boundary data to extract relevant regions in West and Central Africa, focusing on the Lake Chad Basin area.

# Filter to include only countries in the Lake Chad Basin region
# I included Central African Republic for visualization purposes
shp_countries <- west_central_africa |>
  filter(admin0Name %in% c("Nigeria", "Niger", "Chad", "Cameroon", "Central African Republic"))

# Create country codes and format administrative division codes
shp_countries <- shp_countries |>
  mutate(
    country_code = case_when(
      admin0Name == "Nigeria" ~ "NGA",
      admin0Name == "Niger" ~ "NER",
      admin0Name == "Chad" ~ "TCD",
      admin0Name == "Cameroon" ~ "CMR"
    ),
    # Extract and format adm1 and adm2 codes
    adm1_code = str_pad(str_extract(admin1Pcod, "\\d+"), 3, pad = "0"),
    adm2_code = str_pad(str_sub(admin2Pcod, 5, 6), 3, pad = "0"),
    # Create a unique administrative code by combining country and division codes
    adm_code = paste0(country_code, adm1_code, adm2_code)
  ) |>
  # Rename columns for clarity
  rename(
    country = admin0Name,
    adm1 = admin1Name,
    adm2 = admin2Name
  ) |>
  # Select only relevant columns
  select(adm_code, country, adm1, adm2, geometry)

# Identify administrative units within the Lake Chad Basin
shp_countries <- shp_countries |>
  mutate(lake_chad_basin = ifelse(adm_code %in% shp_lake_chad_basin$Rowcacode2,
    "Yes", "No"
  ))

# Subset to retain only the administrative units inside the Lake Chad Basin
shp_lake_chad <- shp_countries |>
  filter(lake_chad_basin == "Yes")

# Compute the bounding box for visualization
bbox <- st_bbox(shp_lake_chad)
x_min <- bbox["xmin"]
x_max <- bbox["xmax"]
y_min <- bbox["ymin"]
y_max <- bbox["ymax"]

# Aggregate administrative level 1 (regions/states) within the Lake Chad Basin
shp_adm1_aggregated <- shp_countries |>
  group_by(adm1, lake_chad_basin) |>
  summarise(geometry = st_union(geometry))

# Keep only regions in the Lake Chad Basin and clean names
shp_adm1_aggregated <- shp_adm1_aggregated |>
  filter(lake_chad_basin == "Yes") |>
  mutate(adm1 = str_replace_all(adm1, "_", " ")) |>
  # Compute centroid coordinates for labeling
  mutate(
    x = st_coordinates(st_centroid(geometry))[, 1],
    y = st_coordinates(st_centroid(geometry))[, 2]
  )

Now, we create a hexagonal grid over the Lake Chad region with a cell size of approximately 10.69 km and fatalities and attack counts are aggregated per grid cell.

# Reproject Lake Chad Basin shapefile to UTM Zone 33N (EPSG:32633)
# This is important to set a cell size in km
shp_lake_chad <- st_transform(shp_lake_chad, crs = 32633)

# Create a hexagonal grid over the Lake Chad Basin with a cell size of ~10.69 km
nc_grid <- st_make_grid(shp_lake_chad,
  cellsize = 10690,
  square = FALSE
)

# Filter the grid to keep only hexagons intersecting the Lake Chad Basin
nc_grid_filter <- nc_grid |>
  st_sf() |>
  st_filter(shp_lake_chad)

# Reproject the grid and Lake Chad Basin back to WGS84 (EPSG:4326)
nc_grid_filter <- st_transform(nc_grid_filter, crs = 4326)
shp_lake_chad <- st_transform(shp_lake_chad, crs = 4326)

# Spatially join conflict data with the hexagonal grid
conflict_data_grid <- st_join(nc_grid_filter, conflict_data_lake_chad, left = FALSE)

# Aggregate conflict data by grid cell, summing fatalities and counting attacks
fatalities_per_grid <- conflict_data_grid |>
  group_by(nc_grid) |>
  summarise(
    fatalities = sum(fatalities, na.rm = TRUE),
    num_attacks = n()
  ) |>
  st_as_sf() |>
  # Compute centroid coordinates for visualization
  mutate(
    x = st_coordinates(st_centroid(nc_grid))[, 1],
    y = st_coordinates(st_centroid(nc_grid))[, 2]
  )

We need to relocate the coordinates of the labels for specific regions for better readability.

# Adjust label positions for specific administrative regions to improve readability
shp_adm1_aggregated <- shp_adm1_aggregated |>
  mutate(
    x = case_when(
      adm1 == "Extreme Nord" ~ x - 0.1, # Shift label slightly left
      TRUE ~ x
    ),
    y = case_when(
      adm1 == "Extreme Nord" ~ y - 0.5, # Move label downward
      adm1 == "Adamawa" ~ y + 0.3, # Move label upward
      adm1 == "Borno" ~ y - 0.4, # Adjust label downward
      TRUE ~ y
    )
  )

# Aggregate country geometries by merging administrative boundaries
# Dissolve some shapes
shp_countries_countries <- shp_countries |>
  group_by(country) |>
  summarise(geometry = st_union(geometry)) |>
  st_cast("POLYGON") |>
  group_by(country) |>
  # Retain the largest polygon for each country
  filter(st_area(geometry) == max(st_area(geometry))) 

# Define manual label positions for country names
country_labels <- tibble(
  country = c("Nigeria", "Niger", "Chad", "Cameroon"),
  x = c(10.15, 11, 15, 13.6), # X-coordinates for label placement
  y = c(10, 15, 14.2, 9) # Y-coordinates for label placement
)

We first create an empty map of the countries and highlight the Lake Chad Basin area using {ggplot2}.

# Automatically use showtext for handling fonts
showtext_auto()

# Add Google font "Voltaire"
font_add_google("Voltaire", "Voltaire")

lake_chad_map <- ggplot() +

  # Plot country borders with a light fill color
  geom_sf(
    data = shp_countries,
    fill = "#ECEFF1",
    color = NA,
    linewidth = 0,
    inherit.aes = FALSE
  ) +

  # Plot Lake Chad Basin area with a darker fill color
  geom_sf(
    data = shp_lake_chad,
    fill = "#E0E0E0",
    color = NA,
    linewidth = 0,
    inherit.aes = FALSE
    )

Now, we add the fatalities information.

lake_chad_map <- lake_chad_map +
  # Plot fatalities per grid with a color gradient
  geom_sf(
    data = fatalities_per_grid,
    aes(fill = fatalities),
    color = NA,
    alpha = 1
  ) +
  scale_fill_gradient(name = "Fatalities", 
                      low = "#FCC5C0", high = "#67001F")

We add different color to subnational and national boundaries and add their corresponding labels.

lake_chad_map <- lake_chad_map +
  # Plot aggregated administrative region boundaries with white lines
  geom_sf(
    data = shp_adm1_aggregated,
    fill = NA,
    color = "white",
    linewidth = 0.5
  ) +

  # Plot country boundaries with black lines
  geom_sf(
    data = shp_countries_countries,
    fill = NA,
    color = "black",
    linewidth = 1.25
  ) +
  # Add labels for administrative regions (e.g., "Extreme Nord")
  geom_text(
    data = shp_adm1_aggregated,
    aes(
      x = x, y = y,
      # Wrap text if needed for long region names
      label = str_wrap(adm1, 1) 
    ),
    colour = "#999",
    fontface = "bold",
    size = 3,
    family = "Voltaire"
  ) +
  
  # Add labels for countries with a larger font
  geom_label(
    data = country_labels,
    aes(x = x, y = y, label = country),
    colour = "black",
    fontface = "bold",
    size = 4,
    family = "Voltaire"
  )

Finally, we center the map at the Lake Chad Basin area and add format for a better visualization.

# Title, subtitle, and caption for the waffle chart
title_chart <- "Armed conflict in the Lake Chad Basin, 2020-2024"
subtitle_chart <- "Fatalities in attacks perpetrated by organized armed actors against civilians"

lake_chad_map <- lake_chad_map +

  labs(
    title = title_chart,
    subtitle = subtitle_chart,
    caption = "**Data:** Armed Conflict Location & Event Data (ACLED) 
 **Graphic:** Juan Torres Munguía",
    x = "",
    y = "",
    fill = ""
  ) +

  # Set the map's coordinate limits and prevent expansion
  coord_sf(
    xlim = c(x_min, x_max),
    ylim = c(y_min, y_max),
    expand = FALSE
  ) +

  # Add a scale bar in the bottom left
  annotation_scale(location = "bl", width_hint = 0.2) +

  # Add a north arrow in the top left corner
  annotation_north_arrow(
    location = "tl",
    which_north = "true",
    style = north_arrow_fancy_orienteering,
    height = unit(1, "cm"),
    width = unit(1, "cm")
  ) +

  # Customize the legend
  guides(
    fill = guide_legend(position = "bottom")
  ) +

  # Set the base theme without any background elements
  theme_void(base_family = "Voltaire") +

  # Customize the map's text and legend styling
  theme(
    # Title settings
    plot.title.position = "plot", # Position of the title
    plot.title = element_textbox(
      color = "black",
      face = "bold",
      size = 22,
      margin = margin(5, 3, 5, 3), # Top, right, bottom, left
      # Width of the title, npc == 1 corresponds to the full width of the plot
      width = unit(1, "npc") 
    ),
    plot.subtitle = element_textbox(
      color = "grey50",
      face = "bold",
      size = 14,
      margin = margin(20, 3, 10, 3),
      width = unit(1, "npc")
    ),

    # Caption settings
    plot.caption = element_markdown(
      color = "grey70",
      size = 10,
      margin = margin(10, 3, 5, 3)
      ),
    legend.title = element_text(hjust = 0.5, color = "black", face = "bold"),
    legend.text = element_text(color = "black", face = "bold"),
    legend.direction = "horizontal",
    legend.title.position = "top",
    legend.text.position = "bottom",
    legend.key.size = unit(0.65, units = "cm"),
    legend.background = element_rect(fill = NA, color = NA),
    legend.margin = margin(t = 5, r = 10, b = 5, l = 10),
    plot.background = element_rect(fill = "#ECEFF1", color = NA)
  )

We create a zoomed-in map of the Lake Chad Basin inset within the African continent, using ggplot2. It includes the ne_countries function from the {rnaturalearth} package to retrieve the map of Africa. The map of the Lake Chad Basin is overlaid on top of Africa, and a rectangle is drawn to highlight the region of interest (Lake Chad Basin). The map is customized with a white background and black borders.

# Fetch African countries map from the rnaturalearth package
africa <- ne_countries(continent = "Africa", returnclass = "sf")

# Create a zoomed-in map of Lake Chad Basin within the African map
lake_chad_map_zoom <- ggplot() +

  # Plot the Africa map with a light fill color
  geom_sf(
    data = africa,
    fill = "#ECEFF1",
    color = NA,
    linewidth = 0,
    inherit.aes = FALSE
  ) +

  # Plot Lake Chad Basin area with a darker fill color
  geom_sf(
    data = shp_lake_chad,
    fill = "#E0E0E0",
    color = NA,
    linewidth = 0,
    inherit.aes = FALSE
  ) +

  # Add a rectangle highlighting the Lake Chad Basin on the Africa map
  geom_rect(
    aes(xmin = x_min, xmax = x_max, ymin = y_min, ymax = y_max),
    color = "black",
    fill = NA,
    linewidth = 0.2
  ) +

  # Remove grid and axes for a clean map appearance
  theme_void() +

  # Customize borders and background of the map
  theme(
    panel.border = element_rect(
      color = "black",
      fill = NA,
      size = 0.8
    ),
    plot.background = element_rect(fill = "white")
  )

Finally, we combine the main map of the Lake Chad Basin with an inset map of Africa showing the location of the Lake Chad Basin. The inset map is positioned within the main map using the inset_element() function of the {patchwork} package. Additionally, we set the resolution of the saved image to 320 dpi (high resolution) for better quality. Finally, the map is saved as a PNG file using ggsave()

# locate main and inset map
lake_chad_map +
  inset_element(
    align_to = "plot",
    lake_chad_map_zoom,
    right = 1,
    left = 0.84,
    top = 1,
    bottom = 0.91,
    on_top = TRUE
  )

# Set the resolution of the image 320 dpi is for high-quality images ("retina")
showtext_opts(dpi = 320) 
ggsave(
  "lake-chad-map.png",
  dpi = 320,
  width = 5,
  height = 12,
  units = "in"
)
showtext_auto(FALSE)

Citation

BibTeX citation:

@online{torres munguía2025,
  author = {Torres Munguía, Juan Armando},
  title = {Visualizing Armed Conflict Using {ACLED’s} Data and {HDX}
    Shapefiles},
  date = {2025-03-09},
  url = {https://juan-torresmunguia.netlify.app/blog/posts/visualizing-conflict-lake-chad-basin},
  langid = {en}
}

For attribution, please cite this work as:

Torres Munguía, Juan Armando. 2025. “Visualizing Armed Conflict Using ACLED’s Data and HDX Shapefiles.” March 9, 2025. https://juan-torresmunguia.netlify.app/blog/posts/visualizing-conflict-lake-chad-basin.

How to compute and visualize the Standardized Precipitation Index (SPI) for drought analysis

Juan Armando Torres Munguía — Fri, 14 Feb 2025 06:00:00 GMT

Overview

Previously, in Part 1 of this tutorial, we calculated the Standardized Precipitation Index (SPI) for different time scales using the {SCI} package in R. Now, we will map the SPI values for the districts of Somalia using {sf} and {ggplot2} packages in R.

About the data

The information of the Somalian sub national administrative boundaries is obtained from the Humanitarian Data Exchange (HDX) initiative of the United Nations Office for the Coordination of Humanitarian Affairs (OCHA). The information comes as a shapefile .shp and is available here in a zip file named som_adm_ocha_20250108_AB_SHP.zip. The shapefile contains the administrative boundaries of Somalia at the subnational level. In particular, we will use file som_admbnda_ADM2_ocha_20250108.shp, that contains the shapefile of Somalia at the district level. I downloaded this file and located it in /shp folder.

Set-up

Before we begin, we need to install and load the necessary R packages.

library(tidyverse) 
library(sf) 

library(ggplot2) 
library(ggtext) 
library(showtext) 
library(patchwork)

Loading data HDX

Data can be found here in file som-rainfall-adm2-full.csv. For Somalia, data is available from January 1981.

shp_somalia <- read_sf("shp/som_admbnda_adm2_ocha_20250108.shp")

shp_somalia <- shp_somalia |>
  rename(
    district = ADM2_EN, 
    district_code = ADM2_PCODE
  )

Data wrangling

I will add the SPI values previously calculated to the shapefile data frame. I saved the SPI values in a RData file named rainfall_data.RData.

load("rainfall_data.RData")

# First, we need to filter the SPI values for the last available month, i.e. December 2024
# to have only one observation per district
rainfall_data_dec2024 <- rainfall_data |>
  filter(month == 2 & year == 2024)

# Merge the shapefile with the SPI values
shp_somalia <- shp_somalia |>
  left_join(rainfall_data_dec2024, by = c("district_code"))

Set theme settings and define fonts, colors, and text to be used in the map

font_add_google("Roboto Condensed", "Roboto Condensed")

showtext_auto()

theme_spi_map <- function() {
  theme_minimal(
    base_family = "Roboto Condensed" 
  ) +
    # Custom theme settings
    theme(

      # Title settings
      plot.title.position = "plot", 
      plot.title = element_textbox(
        color = "black",
        face = "bold",
        size = 22,
        margin = margin(5, 0, 5, 0), 
        width = unit(1, "npc") 
      ),
      plot.subtitle = element_textbox(
        color = "grey50",
        face = "bold",
        size = 14,
        margin = margin(20, 0, 10, 0),
        width = unit(1, "npc")
      ),

      legend.position = "top",
      legend.title = element_blank(),
      legend.key.height = unit(0.1, "cm"), 
      legend.key.width = unit(0.1, "cm"), 
      legend.spacing.x = unit(0.1, "cm"),
      legend.key.spacing = unit(0.1, "cm"), 
      legend.text = element_text(
        margin = margin(5, 0, 5, 0),
        face = "bold",
        color = "grey10",
        size = 10
      ),
      legend.direction = "horizontal",
      legend.byrow = FALSE,

      # Caption settings
      plot.caption = element_markdown(
        color = "grey70",
        face = "italic",
        size = 10,
        hjust = 0
      ),
      plot.background = element_rect(
        color = "white",
        fill = "white"
      ),
      plot.margin = margin(20, 40, 20, 40)
    )
}

# Title, subtitle, and caption for the waffle chart
title_chart <- "Standardized Precipitation Index (SPI) in Somalia"
subtitle_chart <- "February 2024"
caption_chart <- "**Data:** Somalia: Own estimations using Somalia Rainfall Indicators at Subnational Level (HDX) 
 **Graphic:** Juan Torres Munguía"

Now, lets create a histogram of the SPI values for the districts of Somalia to use as a legend for the map.

spi_colors <- data.frame(
  ymin = c(-5, -2, -1.5, -1, 0, 1, 1.5, 2),
  ymax = c(-2, -1.5, -1, 0, 1, 1.5, 2, 5),
  label = factor(
    c(
      "Extreme dry", "Severe dry", "Moderate dry", "Near normal",
      "Near normal", "Moderate wet", "Severe wet", "Extreme wet"
    ),
    levels = c(
      "Extreme dry", "Severe dry", "Moderate dry", "Near normal",
      "Moderate wet", "Severe wet", "Extreme wet"
    )
  ),
  fill = c(
    "#F76D5E", "#FFAD72", "#FFE099", "#E5FFE5",
    "#E5FFE5", "#99EAFF", "#75D3FF", "#3D87FF"
  )
)

shp_somalia <- shp_somalia %>%
  mutate(spi_category = cut(spi1,
    breaks = c(-Inf, spi_colors$ymax),
    labels = spi_colors$label,
    right = TRUE
  )) |> # right = TRUE means that the intervals are closed on the right
  filter(!is.na(spi1))

histogram_spi <- shp_somalia |>
  filter(!is.na(spi_category)) |>
  ggplot(aes(x = spi_category, fill = spi_category)) +
  geom_bar(color = "white") +
  scale_fill_manual(values = setNames(spi_colors$fill, spi_colors$label)) +
  labs(
    title = "",
    x = "",
    y = "Number of districts"
  ) +
  theme_spi_map() +
  theme(
    legend.position = "none",
    axis.title = element_text(
      color = "grey10",
      face = "bold",
      size = 7
    ),
    axis.text = element_text(
      color = "grey10",
      face = "bold",
      size = 7
    ),
    axis.line = element_blank(),
    panel.grid = element_blank()
  ) +
  coord_flip()

# Read the shapefile of Somalia with the interior region borders
shp_somalia_region <- read_sf("shp/som_admbnda_adm1_ocha_20250108.shp")

shp_somalia_region <- st_transform(shp_somalia_region, 3857) # Web Mercator projection
shp_somalia <- st_transform(shp_somalia, 3857)

map_spi <- ggplot() +
  geom_sf(
    data = shp_somalia, aes(fill = spi_category), color = "grey", linewidth = 0.5
  ) +
  scale_fill_manual(values = setNames(spi_colors$fill, spi_colors$label)) +
  geom_sf(data = shp_somalia_region, fill = NA, color = "black", linewidth = 0.9) +
  labs(
    title = title_chart,
    subtitle = subtitle_chart,
    caption = caption_chart,
    x = "",
    y = "",
    fill = ""
  ) +
  theme_spi_map() +
  theme(
    legend.position = "none",
    # Axis settings
    axis.title = element_blank(),
    axis.line = element_blank(),
    axis.text = element_blank(),
    axis.ticks = element_blank(),
    panel.grid = element_blank()
  )

map_spi +
  inset_element(histogram_spi, left = 0.4, bottom = 0, right = 1, top = 0.4, on_top = FALSE)

showtext_opts(dpi = 320) 
ggsave(
  "spi-somalia-map.png",
  dpi = 320,
  width = 12,
  height = 9,
  units = "in"
)
showtext_auto(FALSE)

Citation

BibTeX citation:

@online{torres munguía2025,
  author = {Torres Munguía, Juan Armando},
  title = {How to Compute and Visualize the {Standardized}
    {Precipitation} {Index} {(SPI)} for Drought Analysis},
  date = {2025-02-14},
  url = {https://juan-torresmunguia.netlify.app/blog/posts/spi_somalia_visualization},
  langid = {en}
}

For attribution, please cite this work as:

Torres Munguía, Juan Armando. 2025. “How to Compute and Visualize the Standardized Precipitation Index (SPI) for Drought Analysis.” February 14, 2025. https://juan-torresmunguia.netlify.app/blog/posts/spi_somalia_visualization.

How to compute and visualize the Standardized Precipitation Index (SPI) for drought analysis

Juan Armando Torres Munguía — Sat, 08 Feb 2025 06:00:00 GMT

Overview

The Standardized Precipitation Index (SPI) quantifies how precipitation deviates from the historical average over multiple time scales (e.g., from 1 to 12 months), allowing for consistent monitoring of droughts and floods across regions and time, making it a valuable tool for early warning and risk assessment systems.

In this tutorial, I will walk through the process of calculating the SPI for Somalia at the subnational level using the {SCI} package in R.

About the data

The rainfall dataset used in this tutorial comes from the Humanitarian Data Exchange (HDX) initiative of the United Nations Office for the Coordination of Humanitarian Affairs (OCHA) and provides monthly precipitation data for Somalia’s subnational districts. It is derived from CHIRPS v2 (Climate Hazards Group InfraRed Precipitation with in situ Station data), using satellite imagery combined with ground-based observations.

For each 10-day period (dekad), the dataset includes the following indicators:

rfh: Total rainfall (mm) over 10 days
r1h: 1-month rolling rainfall sum (mm)
r3h: 3-month rolling rainfall sum (mm)
rfh_avg: Long-term average rainfall for the dekad (mm)
r1h_avg: Long-term average for 1-month rolling rainfall (mm)
r3h_avg: Long-term average for 3-month rolling rainfall (mm)
rfq: Rainfall anomaly (%)
r1q: Rainfall anomaly (%)
r3q: 3-month rainfall anomaly (%)

The data is aggregated at the subnational level and follows administrative boundaries defined by the World Food Programme (WFP), with each unit assigned a unique code (Pcode).

Set-up

Before we begin, we need to install and load the necessary R packages. We will be using the {tidyverse} package for data wrangling, as well as the {SCI} package for calculating the SPI.

library(tidyverse) 
library(SCI) # For Standardized Precipitation Index calculations (climate analysis)
library(conflicted) # For resolving conflicts between function names in different packages
library(kableExtra) 

library(ggplot2) 
library(ggtext) 
library(showtext) 

conflicts_prefer(dplyr::select) # Prefer 'select' function from dplyr package
conflicts_prefer(dplyr::lag) # Prefer 'lag' function from dplyr package
conflicts_prefer(dplyr::filter) # Prefer 'filter' function from dplyr package

Warning

Important to use {conflicted} package to specify preferred functions to avoid naming conflicts between {tidyverse} and {SCI} packages.

Loading data HDX

Data can be found here in file som-rainfall-adm2-full.csv. For Somalia, data is available from January 1981.

# Read the rainfall data for Somalia`
# You can also download the data from the HDX platform and read it locally
rainfall_data <- read.csv("https://data.humdata.org/dataset/ed6e1b4b-8094-47e6-bdf7-f6d56fa7abb9/resource/8b333d58-d69e-418c-b5e3-dd86f12eee05/download/som-rainfall-subnat-full.csv")

Data wrangling

We preprocess the data to extract relevant information to calculate the SPI. We extract the year and month from the date column, calculate the rainfall anomaly, and group the data by year, month, and district_code.

rainfall_data <- rainfall_data |>
  slice(-1) |> # Remove the first row (metadata)
  filter(adm_level == 2) |>
  mutate(
    year = year(date), # Extract year from the 'date' column
    month = month(date), # Extract month from the 'date' column
    district_code = PCODE # Rename the 'ADM2_PCODE' column to 'district_code'
  ) |>
  filter(year < 2025) |> # Filter data up to 2024
  rename(
    rainfall_monthly_total = rfh, # Rename the column 'rfh' to 'rainfall_monthly_total'
    rainfall_monthly_lt_average = rfh_avg # Rename the column 'rfh_avg' to 'rainfall_monthly_lt_average'
  ) |>
  mutate(
    rainfall_monthly_total = as.numeric(rainfall_monthly_total), # Convert rainfall total to numeric
    rainfall_monthly_lt_average = as.numeric(rainfall_monthly_lt_average) # Convert long-term average to numeric
  ) |>
  select(year, month, district_code, rainfall_monthly_total, rainfall_monthly_lt_average) # Select relevant columns

# Group by year, month, and district_code, and calculate mean rainfall values by district
rainfall_data <- rainfall_data |>
  group_by(year, month, district_code) |>
  summarise(
    rainfall_monthly_total = mean(rainfall_monthly_total, na.rm = TRUE), # Calculate the mean of rainfall total
    rainfall_monthly_lt_average = mean(rainfall_monthly_lt_average, na.rm = TRUE) # Calculate the mean of long-term average
  ) |>
  ungroup() |>
  # Optionally, you can calculate rainfall anomaly as a percentage of the long-term average
  # rainfall_anomaly = 1, indicates no anomaly
  # < 1 indicates below average
  # > 1 indicates above average
  mutate(
    rainfall_anomaly = rainfall_monthly_total / rainfall_monthly_lt_average
  )

We then calculate the SPI for different time scales (1, 3, 6, 12 months).

rainfall_data <- rainfall_data |>
  group_by(district_code) |> # Group by district_code
  mutate(
    spi1 = transformSCI(
      rainfall_monthly_total,
      first.mon = 1, # Start month for SPI calculation
      obj = fitSCI(rainfall_monthly_total,
        first.mon = 1, # Start month for SPI calculation
        time.scale = 1, # Time scale for SPI calculation
        distr = "gamma", # Distribution for fitting
        p0 = TRUE # Use initial parameter estimates
      )
    ),
    spi3 = transformSCI(
      rainfall_monthly_total,
      first.mon = 1,
      obj = fitSCI(rainfall_monthly_total,
        first.mon = 1,
        time.scale = 3,
        distr = "gamma",
        p0 = TRUE
      )
    ),
    spi6 = transformSCI(
      rainfall_monthly_total,
      first.mon = 1,
      obj = fitSCI(rainfall_monthly_total,
        first.mon = 1,
        time.scale = 6,
        distr = "gamma",
        p0 = TRUE
      )
    ),
    spi12 = transformSCI(
      rainfall_monthly_total,
      first.mon = 1,
      obj = fitSCI(rainfall_monthly_total,
        first.mon = 1,
        time.scale = 12,
        distr = "gamma",
        p0 = TRUE
      )
    )
  ) |>
  ungroup() |>
  arrange(district_code)

The output table looks like this:

year	month	district_code	rainfall_monthly_total	rainfall_monthly_lt_average	rainfall_anomaly	spi1	spi3	spi6	spi12
1981	1	SO1101	0.9417476	1.744876	0.5397218	-1.1043463	NA	NA	NA
1981	2	SO1101	2.0194174	2.027616	0.9959566	0.1537152	NA	NA	NA
1981	3	SO1101	32.8349507	9.696980	3.3861008	2.0730688	2.0600381	NA	NA
1981	4	SO1101	18.3139157	24.722222	0.7407876	-0.3498783	0.9740066	NA	NA
1981	5	SO1101	6.7928803	17.754692	0.3825963	-1.1682128	0.3482195	NA	NA
1981	6	SO1101	2.7443365	4.763215	0.5761522	-1.1812731	-1.5032149	0.1598925	NA
1981	7	SO1101	17.3592230	14.047789	1.2357264	0.7686934	-0.9048239	0.3881687	NA
1981	8	SO1101	34.8381870	26.758146	1.3019657	0.8281462	0.7648482	0.7461252	NA
1981	9	SO1101	29.1682857	19.003020	1.5349290	1.2521098	1.2306838	0.0193027	NA
1981	10	SO1101	4.1974111	8.226213	0.5102483	-0.3816705	0.7653264	0.1559530	NA

Interpretation

Drought events are indicated by negative SPI values, with more negative values indicating more severe drought conditions. Conversely, positive SPI values indicate wetter-than-average conditions. For demonstrating purposes, let’s visualize the SPI values for Laasqoray district in Sanaag region. One efficient way to visualize SPI values is by using a box plot with SPI categories.

font_add_google("Roboto Condensed", "Roboto Condensed")

showtext_auto()

# Custom theme for the chart
theme_boxplot_chart <- function() {
  theme_minimal(
    base_family = "Roboto Condensed" 
  ) +
    # Custom theme settings
    theme(
      # Axis settings
      axis.title = element_text(
        color = "grey10",
        face = "bold",
        size = 14
      ),
      axis.text = element_text(
        color = "grey10",
        face = "bold",
        size = 12
      ),

      # Title settings
      plot.title.position = "plot", 
      plot.title = element_textbox(
        color = "black",
        face = "bold",
        size = 24,
        margin = margin(5, 0, 5, 0), 
        width = unit(1, "npc") 
      ),

      # Legend settings
      legend.position = "top",
      legend.title = element_blank(),
      legend.key.height = unit(0.5, "cm"),
      legend.key.width = unit(0.5, "cm"),
      legend.spacing.x = unit(0.2, "cm"),
      legend.key.spacing = unit(0.5, "cm"), 
      legend.text = element_text(
        margin = margin(5, 0, 5, 0),
        face = "bold",
        color = "grey10",
        size = 10
      ),
      legend.direction = "horizontal",
      legend.byrow = FALSE,

      # Caption settings
      plot.caption = element_markdown(
        color = "grey70",
        face = "italic",
        size = 12,
        hjust = 0
        ),
      plot.background = element_rect(
        color = "white",
        fill = "white"
      ),
      plot.margin = margin(20, 40, 20, 40)
    )
}

title_chart <- "1-month Standardized Precipitation Index (SPI-1) in Laasqoray, Somalia for the period 1981-2024"
caption_chart <- paste0("**Data:** Own estimations using Somalia rainfall Indicators at Subnational Level (HDX)",
                        "
", 
                        "**Graphic:** Juan Torres Munguía")

#
spi_colors <- data.frame(
  ymin = c(-5, -2, -1.5, -1, 0, 1, 1.5, 2),
  ymax = c(-2, -1.5, -1, 0, 1, 1.5, 2, 5),
  label = factor(
    c(
      "Extremely dry", "Severely dry", "Moderately dry", "Near normal",
      "Near normal", "Moderately wet", "Severely wet", "Extremely wet"
    ),
    levels = c(
      "Extremely dry", "Severely dry", "Moderately dry", "Near normal",
      "Moderately wet", "Severely wet", "Extremely wet"
    )
  ),
  fill = c(
    "#F76D5E", "#FFAD72", "#FFE099", "#99FF99",
    "#99FF99", "#99EAFF", "#75D3FF", "#3D87FF"
  )
)

rainfall_data |>
  filter(district_code == "SO1503") |>
  mutate(month = factor(month, levels = 1:12, labels = month.name)) |> # Convert month to factor
  ggplot(aes(x = month, y = spi1)) +

  # Add box plot with SPI categories
  geom_rect(
    data = spi_colors, aes(ymin = ymin, ymax = ymax, xmin = -Inf, xmax = Inf, fill = label),
    alpha = 0.2, inherit.aes = FALSE
  ) +
  geom_boxplot(size = 1, outlier.size = 2, outlier.shape = 16) +
  scale_fill_manual(values = setNames(spi_colors$fill, spi_colors$label)) +
  labs(
    title = title_chart,
    caption = caption_chart,
    x = "",
    y = "SPI-1",
    fill = ""
  ) +
  guides(fill = guide_legend(nrow = 1)) + # Number of rows in the legend
  theme_boxplot_chart()

showtext_opts(dpi = 320) # Set the resolution of the image 320 dpi is for high-quality images ("retina")
ggsave(
  "spi-somalia.png",
  dpi = 320,
  width = 12,
  height = 9,
  units = "in"
)
showtext_auto(FALSE) # Turn off the showtext functionality

Citation

BibTeX citation:

@online{torres munguía2025,
  author = {Torres Munguía, Juan Armando},
  title = {How to Compute and Visualize the {Standardized}
    {Precipitation} {Index} {(SPI)} for Drought Analysis},
  date = {2025-02-08},
  url = {https://juan-torresmunguia.netlify.app/blog/posts/spi_somalia},
  langid = {en}
}

For attribution, please cite this work as:

Torres Munguía, Juan Armando. 2025. “How to Compute and Visualize the Standardized Precipitation Index (SPI) for Drought Analysis.” February 8, 2025. https://juan-torresmunguia.netlify.app/blog/posts/spi_somalia.

Using a tidy approach in R to analyze Mexico’s 2024 ENVIPE

Juan Armando Torres Munguía — Mon, 27 Jan 2025 06:00:00 GMT

Overview

In this article, I will guide you through the process of analyzing Mexico’s 2024 National Survey of Victimization and Perception of Public Safety (Encuesta Nacional de Victimización y Percepción sobre Seguridad Pública, ENVIPE) using the {tidyverse} framework for survey analysis with the {srvyr} package. Specifically, we will calculate the crime prevalence rate per 100,000 inhabitants by state, according to the victim’s sex ( Tasa de prevalencia delictiva por entidad federativa por cada cien mil habitantes, según sexo de la víctima).

Official tables with this indicator can be found here.

By the end of this article, you will be able to use the {tidyverse} to:

Load the survey microdata from INEGI’s website.
Specify the ENVIPE survey design.
Calculate crime prevalence indicators.

About the ENVIPE

ENVIPE is an official survey that provides nationally and state-representative information on household crime victimization and perceptions of public safety in Mexico. Conducted annually since 2011 by the National Institute of Statistics and Geography (INEGI), ENVIPE collects data on the following crimes:

Total vehicle theft (car, pickup truck, or truck).
Theft of vehicle accessories, spare parts, or tools (car, pickup truck, or truck).
Wall graffiti, intentional vehicle damage, and other types of vandalism.
Burglary or attempted burglary.
Kidnapping.
Enforced disappearance.
Homicide.
Robbery or assault on the street or public transportation.

Theft in a manner different from the above.
Bank fraud.
Consumer fraud.
Extortion.
Threats.
Injuries.
Kidnapping.
Harassment or intimidation.
Groping.
Exhibitionism.
Attempted rape.
Sexual rape.
Other crimes.

All survey-related information, including questionnaires, methodology, glossaries, summaries, and microdata, is fully available at INEGI’s website.

Set-up

First, install and load the required packages:

library(tidyverse) # Tidy data analysis framework
library(srvyr) # Tidy syntax for survey data analysis
library(archive) # Extract files from a ZIP archive
library(kableExtra) # Create HTML table

Loading survey microdata from INEGI

Microdata is available here in various formats (CSV, DBF, DTA, SAV, and RData). In this post, we will use the RData files.

load(archive_read(
  archive = "https://www.inegi.org.mx/contenidos/programas/envipe/2024/microdatos/bd_envipe_2024_RData.zip",
  file = "BD_ENVIPE_2024.RData"
))

Tip

Using the {archive} package, we can directly extract the required file from bd_envipe_2024_RData.zip.

This loads the following objects into the global environment:

THogar: This object contains the interview result, the row of the selected respondent, the household factor, and other household characteristics.
TMod_Vic: This object includes information on the crimes suffered by the selected respondent and their household during the reference year, specifically during 2023. Information comes from the victimization module of the survey.
TPer_Vic1: This object includes information related to the perception of public security within the respondent’s geographic area, antisocial behaviors in their immediate surroundings, and changes in habits due to fear of becoming a crime victim in 2023. Additionally, it presents information regarding the performance of authorities and trust in them. In other words, this table covers Sections IV and V of the main questionnaire.
TPer_Vic2: This object consolidates information related to household and personal victimization, particularly focusing on high-impact crimes experienced by the household (such as kidnapping, enforced disappearance, or homicide) or by the selected respondent. This table includes Sections VI and VII of the main questionnaire.
TSDem: This object contains the socio-demographic characteristics of household members, corresponding to Section III of the main questionnaire.
TSVivienda: This object contains the characteristics of the households captured through the questionnaire cover page, as well as the count of household residents and the number of households. In other words, it includes Sections I and II of the main questionnaire.

Calculating crime prevalence rate per 100,000 thousand inhabitants by state, according to the victim’s sex

To estimate the crime prevalence rate, we use information from TPer_Vic2 (screening questions on victimization from the main questionnaire) and TMod_Vic (specific victimization module applied to the respondents who reported being a victim of at least one of the crimes included in the survey during the reference period). TPer_Vic2 will serve to characterize the total population and TMod_Vic to describe the victims and the associated crimes.

1. Specifying the ENVIPE survey design

ENVIPE follows a probabilistic, three-stage, stratified, and clustered sampling design. The sampling unit is the household, while the unit of analysis is the population aged 18 and older in selected households. More details on the sampling design can be found here(in Spanish).

From the documentation on data structure(in Spanish), the key variables for specifying the survey design are:

ID_PER: Person identifier
EST_DIS: Stratum
UPM_DIS: Primary sampling unit
FAC_ELE: Weighting factor for public security perception and victimization estimates.

We specify the survey design as follows:

# Identify victims from TMod_Vic, excluding
# "Wall graffiti, intentional vehicle damage, and other vandalism" (BPCOD != "03")
id_victims <- TMod_Vic |>
  filter(BPCOD != "03") |>
  pull(ID_PER)

# Create victimization variable in TPer_Vic2
TPer_Vic2 <- TPer_Vic2 |>
  mutate(victimization = ifelse(ID_PER %in% id_victims, 1, 0))

# Convert FAC_ELE (weights) to numeric
TPer_Vic2 <- TPer_Vic2 |>
  mutate(FAC_ELE = as.numeric(FAC_ELE))

# Label SEXO variable
TPer_Vic2 <- TPer_Vic2 |>
  mutate(SEXO = recode(SEXO, "1" = "Man", "2" = "Woman"))

# Create survey object
envipe_design <- TPer_Vic2 |>
  as_survey_design(
    weights = FAC_ELE, # Weights (inverse probability)
    strata = EST_DIS, # Stratum
    ids = UPM_DIS, # Primary sampling unit
    nest = TRUE # Nesting within strata
  )

2. Calculating point estimates and variablity indicators

To compute proportions from survey data, we use survey_mean() from {srvyr} package. From ENVIPE’s documentation, we know that NOM_ENT represents the state, while SEXO corresponds to the respondent’s sex.

envipe_design |>
  # `interact` function is used to calculate the proportion
  # over the interaction of state and sex
  group_by(interact(NOM_ENT, SEXO)) |>
  summarize(
    Proportion = survey_mean(victimization,
      level = 0.90, # Confidence level
      vartype = c("se", "ci", "cv") # Variability indicators
    )
  ) |>
  ungroup() |>
  mutate(
    `Prevalence rate` = Proportion * 100000, # Rate per 100,000 people
    `Standard error` = Proportion_se * 100000,
    `Low 90% CI` = Proportion_low * 100000,
    `Upper 90% CI` = Proportion_upp * 100000,
    `Coefficient of variation %` = Proportion_cv * 100,
    State = str_to_sentence(NOM_ENT) # Capitalize only the first letter
  ) |>
  rename(Sex = SEXO) |>
  select(
    State, Sex, `Prevalence rate`, `Standard error`,
    `Low 90% CI`, `Upper 90% CI`, `Coefficient of variation %`
  ) |>
  kbl() |> # HTML table
  kable_styling() # Apply bootstrap theme to the table

The output table looks like this:

State	Sex	Prevalence rate	Standard error	Low 90% CI	Upper 90% CI	Coefficient of variation %
Aguascalientes	Man	33710.81	2172.7372	30136.71	37284.91	6.445224
Aguascalientes	Woman	32041.57	1831.7081	29028.45	35054.68	5.716662
Baja California	Man	23043.35	1604.9769	20403.20	25683.50	6.965032
Baja California	Woman	24480.64	1536.1771	21953.66	27007.61	6.275070
Baja California Sur	Man	19329.94	1244.1353	17283.36	21376.51	6.436314
Baja California Sur	Woman	23817.58	1466.2963	21405.56	26229.60	6.156361
Campeche	Man	21832.87	1455.8508	19438.03	24227.71	6.668161
Campeche	Woman	23213.74	1381.0933	20941.88	25485.61	5.949465
Chiapas	Man	16602.99	986.0052	14981.03	18224.94	5.938722
Chiapas	Woman	12074.36	978.7862	10464.28	13684.44	8.106318
Chihuahua	Man	22303.22	1234.5065	20272.49	24333.96	5.535103
Chihuahua	Woman	22804.31	1181.0327	20861.54	24747.08	5.178990
Ciudad De Mexico	Man	34562.66	1089.2714	32770.84	36354.49	3.151584
Ciudad De Mexico	Woman	30746.76	990.3528	29117.66	32375.87	3.220999
Coahuila De Zaragoza	Man	18179.70	1249.1422	16124.89	20234.51	6.871084
Coahuila De Zaragoza	Woman	20447.19	1238.7111	18409.54	22484.84	6.058099
Colima	Man	21106.85	1593.5970	18485.43	23728.28	7.550140
Colima	Woman	21295.28	1351.1455	19072.68	23517.88	6.344813
Durango	Man	15240.74	1133.1868	13376.67	17104.80	7.435250
Durango	Woman	17401.18	1166.0910	15482.99	19319.37	6.701219
Guanajuato	Man	20057.38	1428.0981	17708.20	22406.57	7.120063
Guanajuato	Woman	18930.15	1287.0811	16812.93	21047.37	6.799107
Guerrero	Man	16981.89	1494.6554	14523.22	19440.56	8.801466
Guerrero	Woman	14572.44	1238.6234	12534.93	16609.94	8.499769
Hidalgo	Man	22397.82	1338.0474	20196.77	24598.88	5.974006
Hidalgo	Woman	20215.02	1157.0294	18311.73	22118.30	5.723613
Jalisco	Man	24390.13	1548.7916	21842.40	26937.85	6.350076
Jalisco	Woman	25141.90	1438.5170	22775.58	27508.23	5.721592
Mexico	Man	32906.59	1632.3183	30221.47	35591.71	4.960460
Mexico	Woman	33023.78	1483.3471	30583.71	35463.85	4.491755
Michoacan De Ocampo	Man	14405.17	1031.1125	12709.01	16101.32	7.157934
Michoacan De Ocampo	Woman	15485.20	1036.3320	13780.46	17189.94	6.692401
Morelos	Man	26995.43	1517.6054	24499.00	29491.85	5.621713
Morelos	Woman	24869.74	1688.5969	22092.04	27647.44	6.789765
Nayarit	Man	16075.56	1516.6195	13580.76	18570.36	9.434318
Nayarit	Woman	18639.61	1597.1842	16012.28	21266.94	8.568766
Nuevo Leon	Man	23602.12	1261.9498	21526.25	25678.00	5.346764
Nuevo Leon	Woman	21440.72	1211.6372	19447.60	23433.83	5.651104
Oaxaca	Man	12940.56	1158.2366	11035.29	14845.83	8.950434
Oaxaca	Woman	13548.61	1017.7020	11874.51	15222.70	7.511489
Puebla	Man	26519.64	1342.0473	24312.00	28727.27	5.060579
Puebla	Woman	24176.45	1124.4656	22326.73	26026.17	4.651078
Queretaro	Man	30297.50	1620.6534	27631.57	32963.44	5.349132
Queretaro	Woman	25746.31	1260.3524	23673.06	27819.56	4.895273
Quintana Roo	Man	22592.04	1384.4250	20314.70	24869.39	6.127932
Quintana Roo	Woman	21775.64	1237.8095	19739.48	23811.81	5.684376
San Luis Potosi	Man	24019.02	1874.7438	20935.11	27102.93	7.805247
San Luis Potosi	Woman	23227.97	1491.3111	20774.80	25681.14	6.420324
Sinaloa	Man	21181.44	1122.0160	19335.76	23027.13	5.297165
Sinaloa	Woman	20562.22	1086.2790	18775.32	22349.12	5.282887
Sonora	Man	26864.66	1816.2677	23876.94	29852.37	6.760808
Sonora	Woman	25742.16	1902.0264	22613.37	28870.95	7.388759
Tabasco	Man	25881.72	1354.6212	23653.40	28110.04	5.233891
Tabasco	Woman	25522.45	1091.8488	23726.38	27318.51	4.277994
Tamaulipas	Man	17788.15	1075.8038	16018.47	19557.82	6.047869
Tamaulipas	Woman	18368.67	1017.1639	16695.46	20041.88	5.537493
Tlaxcala	Man	23968.33	1621.2198	21301.47	26635.20	6.764007
Tlaxcala	Woman	22295.82	1473.2391	19872.38	24719.26	6.607692
Veracruz De Ignacio De La Llave	Man	16365.05	1055.8247	14628.24	18101.86	6.451705
Veracruz De Ignacio De La Llave	Woman	14712.45	947.9121	13153.15	16271.74	6.442927
Yucatan	Man	20677.82	1239.0909	18639.55	22716.10	5.992366
Yucatan	Woman	17805.62	1141.8291	15927.34	19683.90	6.412745
Zacatecas	Man	17328.49	1346.8466	15112.96	19544.02	7.772440
Zacatecas	Woman	15690.25	1335.9463	13492.65	17887.85	8.514502

Citation

BibTeX citation:

@online{torres munguía2025,
  author = {Torres Munguía, Juan Armando},
  title = {Using a Tidy Approach in {R} to Analyze {Mexico’s} 2024
    {ENVIPE}},
  date = {2025-01-27},
  url = {https://juan-torresmunguia.netlify.app/blog/posts/estimating-envipe-prevalence},
  langid = {en}
}

For attribution, please cite this work as:

Torres Munguía, Juan Armando. 2025. “Using a Tidy Approach in R to Analyze Mexico’s 2024 ENVIPE.” January 27, 2025. https://juan-torresmunguia.netlify.app/blog/posts/estimating-envipe-prevalence.

A waffle chart of disease outbreaks in the world

Juan Armando Torres Munguía — Sun, 12 Jan 2025 06:00:00 GMT

Overview

In these notes, I will show you how to use {ggplot} to create a waffle chart about the frequency of disease outbreaks in the world. The final output will look like this:

About the data

The source of information is the global dataset of pandemic- and epidemic-prone disease outbreaks, whose data is freely available at the GitHub repository of the disease outbreaks project.

The unit of analysis in the database is an outbreak, defined as the occurrence of at least one case of a specific disease in a country -or territory- during a particular year. Therefore, a country -or territory- cannot have more than one outbreak of the same disease in the same year, although it may experience outbreaks of different diseases within the same year. A country can only have multiple outbreaks of the same disease if they occur in different years.

The dataset includes information on over 3,000 pandemic- and epidemic-prone disease outbreaks associated with more than 80 different infectious diseases, occurring globally from January 1996 to the present. The diseases are classified according to the International Classification of Diseases, 10th Revision (ICD-10), and the dataset contains information on the year, country, and disease for each outbreak.

Set-up

To create the waffle chart, we will use the following R packages:

library(tidyverse) # Tidy data analysis framework
library(readxl) # Read Excel files
library(httr) # Perform HTTP requests
library(showtext) # Add custom fonts to plots
library(ggtext) # Use rich text elements in ggplot
library(waffle) # Create waffle charts

Loading data from GitHub

The dataset is stored in an Excel. We will download it from the GitHub repository and load it into R:

# Define the GitHub raw content URL
link_data <- "https://github.com/jatorresmunguia/disease_outbreak_news/raw/refs/heads/main/Humanitarian%20Data%20Exchange/data_2_share/disease_outbreaks_HDX.xlsx"

# Create a temporary file to store the data
temp <- tempfile(fileext = ".xlsx")

# Download the data from GitHub using authentication
get_req <- GET(
  link_data,
  authenticate(Sys.getenv("GITHUB_PAT"), ""),
  write_disk(path = temp)
)

# Read the Excel file into R
outbreaks <- read_excel(temp)

Data wrangling

We will group the data by year and disease, and then filter the top 10 diseases from 1996 to 2024. We will also group the data by year and disease, and then filter the top 10 diseases from 2015 to 2024. Finally, we will create a factor variable for the top 10 diseases and the “Other” category:

# Top 10 diseases 1996-2024
top_10_icd104n <- outbreaks |>
  filter(Year < 2025) |>
  group_by(icd104n) |>
  summarise(freq = n(), .groups = "drop") |>
  arrange(-freq) |>
  slice_head(n = 10)

# Only top-10 categories and Others
outbreaks_year_disease_grouped <- outbreaks |>
  filter(Year < 2025 & Year > 2014) |>
  mutate(icd104n = ifelse(icd104n %in% top_10_icd104n$icd104n, icd104n, "Other")) |> # Group the diseases
  group_by(Year, icd104n) |>
  summarise(freq = n(), .groups = "drop") |> # Count the number of outbreaks
  arrange(-freq)

# Create a factor variable for the top 10 diseases and the "Other" category
outbreaks_year_disease_grouped <- outbreaks_year_disease_grouped |>
  mutate(icd104n = factor(icd104n,
    levels = c(
      "Acute poliomyelitis, unspecified", "Classical cholera", "COVID-19, virus identified",
      "Dengue, unspecified", "Influenza due to identified zoonotic or pandemic influenza virus", "Measles",
      "Meningococcal meningitis", "Middle East respiratory syndrome coronavirus [MERS-CoV]", "Monkeypox",
      "Yellow fever, unspecified", "Other"
    ),
    labels = c(
      "Acute poliomyelitis", "Cholera", "COVID-19", "Dengue", "Influenza",
      "Measles", "Meningococcal meningitis", "MERS-CoV", "Monkeypox",
      "Yellow fever", "Other"
    )
  ))

Set theme settings and define fonts, colors, and text to be used in the waffle chart

# Add custom font
font_add_google("Roboto Condensed", "Roboto Condensed") 

showtext_auto()

# Custom theme for the waffle chart
theme_waffle_chart <- function() {
  theme_minimal(
    base_family = "Roboto Condensed" # Base theme with custom font
  ) +
    # Custom theme settings
    theme(
      # Axis settings
      axis.title = element_blank(), # Remove axis title
      axis.line = element_blank(), # Remove axis line
      
      # Title settings
      plot.title.position = "plot", # Position of the title
      plot.title = element_textbox(
        color = "black",
        face = "bold",
        size = 24,
        margin = margin(5, 0, 5, 0), # Top, right, bottom, left
        width = unit(1, "npc") # Width of the title, npc == 1 corresponds to the full width of the plot
      ),
      
      # Subtitle settings
      plot.subtitle = element_textbox(
        color = "grey50",
        face = "bold",
        size = 14,
        margin = margin(0, 0, 10, 0),
        width = unit(1, "npc")
      ),
      
      # Legend settings
      legend.position = "top",
      legend.title = element_blank(),
      legend.spacing.x = unit(0.2, "cm"),
      legend.key.spacing = unit(0.5, "cm"), # Spacing between legend keys
      legend.text = element_text(
        margin = margin(5, 2, 5, 0),
        face = "bold",
        color = "grey10",
        size = 9
      ),
      legend.direction = "horizontal",
      legend.byrow = FALSE,
      
      # Caption settings
      plot.caption = element_markdown(
        color = "grey70",
        size = 8,
        hjust = 0
          ),
      plot.caption.position = "plot",

      plot.background = element_rect(
        color = "white",
        fill = "white"
      ),
      plot.margin = margin(20, 40, 20, 40)
    )
}

# Title, subtitle, and caption for the waffle chart
title_chart <- "Pandemic- and epidemic-prone disease outbreaks in the world"
subtitle_chart <- "In 2024, fewer countries reported public health events linked to COVID-19, while outbreaks of viral diseases transmitted by vectors increased"
caption_chart <- paste0("**Note:** Each square represents a country.",
                        "
", 
                        "**Data:** A global dataset of pandemic- and epidemic-prone disease outbreaks (DOI: 10.1038/s41597-022-01797-2)",
                        "
", 
                        "**Graphic:** Juan Torres Munguía"
                        )

# Define the colors to be used in the waffle chart
colors <- c(
  "#B66DFF", "#7e0021", "#ff420e", "#579d1c", "#83caff",
  "#0084d1", "#aecf00", "#4b1f6f", "#FF86FF", "#ffd320", "grey80"
  )

Create the waffle chart

ggplot(
  outbreaks_year_disease_grouped,
  aes(fill = icd104n, values = freq)
) +
  geom_waffle(
    size = 0.25, # Size of the waffle squares
    n_rows = 10, # Number of rows in the waffle chart
    flip = TRUE, # Flip the waffle chart
    color = "white", # Color of the border of the waffle squares
    make_proportional = FALSE
  ) + # Do not make the waffle chart proportional
  facet_wrap(~Year, # Facet the waffle chart by year
             nrow = 1,
             strip.position = "bottom"
  ) + # Position of the facet strip
  scale_fill_manual(values = colors) +
  scale_x_discrete() +
  scale_y_continuous(
    labels = function(x) x * 10,
    expand = c(0, 0)
  ) + # Expand the y-axis
  coord_equal() +
  labs(
    title = title_chart,
    subtitle = subtitle_chart,
    caption = caption_chart,
    x = "",
    y = "Total number of outbreaks",
    fill = ""
  ) +
  guides(fill = guide_legend(nrow = 2 # Number of rows in the legend
                             )) + 
  theme_waffle_chart()

Save the waffle chart as an image

showtext_opts(dpi = 320) # Set the resolution of the image 320 dpi is for high-quality images ("retina")
ggsave(
  "waffle-pandemics.png",
  dpi = 320,
  width = 14,
  height = 7,
  units = "in"
)
showtext_auto(FALSE)

Citation

BibTeX citation:

@online{torres munguía2025,
  author = {Torres Munguía, Juan Armando},
  title = {A Waffle Chart of Disease Outbreaks in the World},
  date = {2025-01-12},
  url = {https://juan-torresmunguia.netlify.app/blog/posts/waffle-chart-disease-outbreaks},
  langid = {en}
}

For attribution, please cite this work as:

Torres Munguía, Juan Armando. 2025. “A Waffle Chart of Disease Outbreaks in the World.” January 12, 2025. https://juan-torresmunguia.netlify.app/blog/posts/waffle-chart-disease-outbreaks.

Data Art & Science Blog

How to build a waffle chart with circle-shaped tiles using {waffle} and {ggplot2} libraries in R?

Overview

About the data

Set-up

Loading data

Step 1. Visual elements of the plot

Step 2. Designing the waffle plot using ggplot2 package

Step 3. Transforming the squares into circles.

Step 4. Save the waffle chart as a high-quality image

References

Citation

Designing a map to visualize the geographic distribution of pandemic- and epidemic-prone disease outbreaks in 2025

Overview

About the data

Set-up

Loading data

Step 1. Data wrangling

Step 2. Geographic information

Identifying continents

Step 3. Enhancing other visual elements of the plot

Step 4. Putting together all the previously designed elements

References

Citation

How has the number of persons seeking protection in Germany changed over time?

Overview

About the data

Set-up

Loading data

Designing the stacked area chart

Citation

Creating a population pyramid to visualize crime victimization by age and sex in Mexico

Overview

Set-up

Loading data

Creating the population pyramid

Citation

A dot plot of public acceptability of domestic violence in Cambodia between 2005 and 2009

Overview

Set-up

Loading data

Dot plot design

Citation

Three maps to examine the prevalence of intimate partner violence (IPV) in Cambodia

Overview

About the data

Set-up

Loading data

Data wrangling

Creating the map

Citation

Working with Humanitarian Exchange Language (HXL) Hashtags

Setting Up the Environment

Step 1: Accessing the Dataset

Step 2: Adding WHO and UNSD Regional Classifications

WHO Regions

UNSD Regions

Step 3: Structuring the Dataset

Step 4: Adding HXL Hashtags

Step 5: Exporting the Final Dataset

Citation

Visualizing armed conflict using ACLED’s data and HDX shapefiles

Overview

Set-up

Loading data

Data processing

Citation

How to compute and visualize the Standardized Precipitation Index (SPI) for drought analysis

Overview

About the data

Set-up

Loading data HDX

Data wrangling

Set theme settings and define fonts, colors, and text to be used in the map

Citation

How to compute and visualize the Standardized Precipitation Index (SPI) for drought analysis

Overview

About the data

Set-up

Loading data HDX

Step 2. Designing the waffle plot using `ggplot2` package