DataVis

Here is a collection of some artifacts produced during my time thinking about and tinkering with data visualisations. The inspiration comes from my admiration for artists linked to Latin American geometric abstraction, especially Joaquín Torres García, Lygia Clark and Loló Soldevilla.

Co-laborators

Almost all my publications (reports, chapters, journal articles) result from collaborations with dear colleagues. These two plots provide a visual representation of the duration and diversity of this co-laboured trajectory.

Composição

1. First, we import the dataset that contains basically the surnames of my colleagues, their country of origin and preferred pronoun, as well as the year we started working together.

2. Using the package imgpalr, we create an object with the HEX code for colours that are present in random points of Lygia Clark’s painting “Composição (1953)”.

3. Finally, we generate one rectangle for each colleague with the side proportional to the time since we met. Colours represent the combination of pronoun and country of origin.

library(tidyverse)
library(ggpubr)
library(imgpalr)

# 1) Importing the dataset  and transforming variables ########################

# Reading the dataset
df <- read_csv("./data/colleagues.csv")

# Getting current year
this_year <- as.numeric(format(Sys.Date(), "%Y"))

# Size of the block is equal to the number of years since year met
df$size <- (this_year - df$year_met)

# Creating group combining country and pronoum
df$group <- paste0(df$country, df$pronoum)

# Number of co-laborators
n_colab <- nrow(df)


# 2) Creating colour palette #####################################

# Seed for palette generation and coordinates
seed = 7
set.seed(seed)

# URL of the image of Lygia Clark's Composicao (1953)

img_url <- "https://portal.lygiaclark.org.br/public/upload/screen/2021-09-07/6b1143a6023bc122578da3423f4149cc[2100x2995].jpg"

# Generating one color for each row in the dataset
compo <- image_pal(img_url, n = nrow(df), type = "qual", seed = seed)


# 3) Generating the plot #####################################

# Adding empty columns for coordinates
df <- cbind(df, x1 = NA, x2 = NA, y1 = NA, y2 = NA)

# Loop to generate random coordinates that overlap slightly
for (i in sample(1:n_colab)){
  # Defining the starting point randomly
  df$x1[i] <- sample(1:n_colab)
  df$y1[i] <- sample(1:n_colab)
  
  # If pronoun equals to s then go right and up otherwise the opposite
  direction <- if_else(df$pronoum[i] == "s", 1, -1)
  height <- sample(c(0.1, 0.7, 1), 1)

  df$x2[i] <- df$x1[i] + direction * df$size[i]
  df$y2[i] <- df$y1[i] + direction * df$size[i] * height
}

plot_compo <- 
  ggplot(df,aes(xmin = x1, xmax = x2, ymin = y1, ymax = y2)) +
    geom_rect(aes(fill = group, alpha = 1 - size/max(size))) +
    scale_fill_manual(values = compo) +
    theme_void() +
    theme(legend.position = "none")

ggsave(plot_compo, filename = "./_img/composicao.png", width = 8, height = 4)

plot_compo + 
  theme(panel.border = element_rect(colour = "grey30", fill=NA, linewidth=1)) 

Composicao (1953) - Lygia Clark (Source: Associação Cultural Lygia Clark)

Timeline

The second plot follows the same idea of representing duration and diversity but takes a simpler approach of plotting the timeline as coloured line segments and then “bending” the plot with the function coord_plot(). The composition below displays the same plot with different starting points.

# Y random position for segments
df$pos <- sample(1:n_colab, n_colab)

# Generating the ggplot
plot <- 
  ggplot(df, aes(x = year_met, xend = this_year,
                  y = pos, yend = pos)) +
  geom_segment(aes(color = group), linewidth = 1.2) +
  scale_color_manual(values = compo) +
  theme_void() +
  xlim(1985, this_year) +
  theme(legend.position = "none",
        plot.background = element_rect(fill = "#F5F5DD", 
                          colour = "white", linewidth = 5),
        plot.margin = margin(20, 2, 1, 2)) 

plot_time <- 
  ggarrange(
    plot + coord_polar(start = 0.3),
    plot + coord_polar(start = 0.6),
    plot + coord_polar(start = 0.9),
    nrow = 1
  )

ggsave(plot_time, filename = "./_img/timeline.png")

plot_time

UIS map

In 2019, I worked in the elaboration of the SDG4 Data Digest published by the UNESCO Institute for Statistics (UIS). Considering the importance of having SDG4 data available for all countries, Sylvia Montoya and I decided to represent each country equally to display data availability in the publication.

This visualisation is based on a modified version of the “World Tile Grid Map” developed by Jonathan Schwabish with contributions from Maarten Lambrechts.

The colours below represent each UIS region, and the squares were filled with colour only if the country had data available. For indicators with low coverage, most of the map was greyed out, like the map for the gender parity index of the indicator 4.1.1b, shown on the sidebar.

#Required
library(tidyverse)
library(countrycode)

#Loading template map####

#World Tile Grid (WTG)
#Loading plot coord (source: "World Tile Grid Map" elaborated by Maarten Lambrechts)
#countries included manually: Palestine, Cook Islands, Andorra (UNESCO member states)
#countries removed manually: Antarctica, Kosovo, Greenland

tb_wtg <- read_csv("./data/uis_worldtilegrid.csv")
tb_wtg$alpha.2 <- as.character(tb_wtg$alpha.2)
tb_wtg$alpha.3 <- as.character(tb_wtg$alpha.3)

# Including UIS SDG regions
#Source https://unstats.un.org/sdgs/indicators/regional-groups/  Access 14/06/2019

regions <- read_csv("./data/uis_regions.csv")
regions$alpha.2 <- countrycode(regions$Country, "country.name", "iso2c", nomatch = NULL)
regions$alpha.3 <- countrycode(regions$Country, "country.name", "iso3c", nomatch = NULL)

#joining SDG regions and countries' coords
WTG <- tb_wtg %>% left_join(regions, by = "alpha.3")

#map with all countries
uis_map <- 
  WTG %>% 
    mutate(Region = case_when(Region == "Europe and Northern America" ~ "Europe and North America", 
                              TRUE ~ Region)) %>% 
    ggplot(aes(xmin = x, ymin = y, xmax = x + 1, ymax = y + 1, fill = as.factor(Region))) +
    geom_tile(aes(x= x, y = y), colour = "grey80") + 
    theme_minimal() +
    theme(panel.grid = element_blank(), axis.text = element_blank(), 
          axis.title = element_blank(), legend.position = "none") +
    geom_text(aes(x = x, y = y, label = alpha.2.y), color = "grey20", size = 3) +
    scale_y_reverse() +
    scale_fill_manual(values = c("#ff0000", "darkorange", "#ffef00", "#21a328", 
                                "#0072b8", "#7f6999", "#ee82ee", "grey80"), drop = FALSE,
                    name = "SDG Region") +
    coord_equal() 

ggsave(uis_map, filename = "./_img/uis_map.png", dpi = 320, width = 6, height = 5, bg = "transparent")

uis_map

Figure 16 of the SDG4 Data Digest 2019. Source: UIS

Palm trees

In 2022, Seaneen Sloan and I worked on the report for the Safe Learning Study. The study assessed a school-based programme in Sierra Leone.

The plot below represents the expected added value in a literacy test comparing students who participated in three different arms of the programme and the control group. The visualisation is inspired by the beautiful dandelion plots by Martin Devaux.

library(tidyverse)

df_leafs <- read_csv("./data/sls_results.csv")

trunk_height <- 13 #Intercept

trunk_pos1 <- 20

df_tree <- 
df_leafs %>% 
  mutate(Group = case_when(
    Group == "Control" ~ trunk_pos1,
    Group == "Group 1" ~ 2*trunk_pos1,
    Group == "Group 2" ~ 3*trunk_pos1,
    Group == "Group 3" ~ 4*trunk_pos1)) %>% 
  mutate(leaf_leng = if_else(Group == trunk_pos1, 1, Estimate + 1),
         leaf_x = case_when(
           outcome == "Letter Names" ~ Group - leaf_leng,
           outcome == "Letter Sounds" ~ Group - leaf_leng,
           outcome == "Invented words" ~ Group + leaf_leng,
           outcome == "Familiar words" ~ Group + leaf_leng,
           outcome == "Oral reading" ~ Group + leaf_leng),
         leaf_y = case_when(
           outcome == "Letter Names" ~ trunk_height - leaf_leng,
           outcome == "Letter Sounds" ~ trunk_height,
           outcome == "Invented words" ~ trunk_height + leaf_leng,
           outcome == "Familiar words" ~ trunk_height,
           outcome == "Oral reading" ~ trunk_height - leaf_leng),
         leaf_pos = case_when(
           outcome == "Letter Names" ~ "left",
           outcome == "Letter Sounds" ~ "left",
           outcome == "Invented words" ~ "right",
           outcome == "Familiar words" ~ "right",
           outcome == "Oral reading" ~ "right"))


mytheme <- 
  theme(
    axis.text.x = element_text(size = 18),
    legend.text = element_text(size = 11),
    plot.background = element_rect(fill = '#f0ebe7', colour = NA),
    plot.margin = margin(15, 80, 15, 80),
    panel.background = element_rect(fill = '#f0ebe7', colour = NA))

gg_palmtree <-
  df_tree %>% 
    ggplot(aes(x =  Group, xend = leaf_x, y = trunk_height, yend = leaf_y, 
              color = outcome))+
    geom_curve(aes(x = Group, xend = Group, y = 0, yend = trunk_height),
              curvature = 0.2, size = 1.5,
              color = '#5F5D36', lineend = "round")+
    geom_curve(data = subset(df_tree,leaf_pos == "left"),
              curvature = 0.6, lineend = "round", size = 4) +
    geom_curve(data = subset(df_tree,leaf_pos == "right"), 
              curvature = -0.6, lineend = "round", size = 4) +
    scale_x_continuous(breaks = c(trunk_pos1, 2*trunk_pos1, 3*trunk_pos1, 4*trunk_pos1),
                      labels = c("Control", "Group 1", "Group 2", "Group 3"))+
    scale_color_brewer(type = "qual", palette = "Dark2",
                      breaks=c("Letter Names","Letter Sounds",
                                  "Invented words", "Familiar words", 
                                  "Oral reading"))+
    ylim(0, 20) +
    theme_void() +
    theme(legend.position = "top",
          legend.title = element_blank()) +
    mytheme

gg_palmtree

ESS and the SDGs