q1.1 Create agooglesheet object from the Google Sheets data set found at data_url.

sheet <- gs_url(x = data_url)

q1.2 The original data is far from tidy. Read in the names, start dates, and end dates of all the atrocities. These can be found in the third quadrant. We only want the first, second-to-last, and last columns. Call the resulting tibble atrocities.

atrocities <- 
  sheet %>% 
  gs_read(range = cell_cols(cols_atrocities), skip = 18, col_names = FALSE) %>% 
  dplyr::select(atrocity = 1, 
         start_year = 13,
         end_year = 14)

q1.3 Now, read in the fourth quadrant up until (and including) column AP. Each column represents a 100-year period of time. The values represent the number of deaths due to the atrocity listed in column A. Set col_names equal to columns and keep verbose equal to TRUE (Don't worry--the column names won't stay numbers for long). Call this tibble deaths.

deaths <- gs_read(sheet, range = cell_cols(cols_deaths), 
                  skip = 18, col_names = column_names)

q1.4 Inspect each tibble and look for anything unusual. Fix any problems you see.

deaths <-
  deaths %>% 
  mutate(`1600` = ifelse(typeof(`1600`) == "character", 0, `1600`))

q1.5 Bind atrocities and deaths. Tidy the data and name the resulting tibble atrocities_tidy. Create a variable type that represents the type of an atrocity (you can use the vectors of types given in the parameters section).

Note: You're probably going to end up with a lot of rows where the number of deaths is 0. I filtered them out so that there would be fewer rows and the tibble would be more interpretable, but it's not strictly necessary.

find_type <- function(atrocity) {
  if (atrocity %in% colonial_wars) {return("colonial war")}
  if (atrocity %in% international_wars) {return("international war")} 
  if (atrocity %in% oppression) {return("institutional oppression")}
  if (atrocity %in% failed_states) {return("failed states")}
  if (atrocity %in% civil_wars) {return("civil war")}
  if (atrocity %in% despots) {return("despots")}
  else {return("NA")}
}

atrocities_tidy <- 
  atrocities %>% 
  bind_cols(deaths) %>% 
  gather(key = "century", value = "deaths", 4:28, convert = TRUE) %>% 
  filter(!(near(deaths, 0))) %>% 
  mutate(type = map_chr(atrocity, find_type))

Atrocities

q2 Let's look at atrocities_tidy. Construct a couple plots that visualize the information. Look at the outliers. What can you conclude?

atrocities_tidy <-
  atrocities_tidy %>% 
  mutate(length = end_year - start_year,
         type = str_to_title(type)) 

to_label <-
  atrocities_tidy %>% 
  group_by(type) %>% 
  distinct(atrocity, start_year, length) %>% 
  top_n(length, n = 2)
  
  
atrocities_tidy %>% 
  group_by(atrocity) %>% 
  ggplot(aes(start_year, length)) +
  geom_point() +
  geom_smooth(method = loess, se = FALSE) +
  geom_label_repel(aes(label = atrocity), size = 2, color = "black", 
                   nudge_y = 25, data = to_label) +
  labs(x = "Starting year",
       y = "Length in years",
       title = "Atrocity length in years by starting year") +
  facet_wrap(~type) +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 45))

atrocities_top <-
  atrocities_tidy %>% 
  group_by(atrocity, start_year, end_year, length, type, century) %>% 
  summarise(total = sum(deaths)) %>% 
  group_by(century) %>% 
  filter(near(total, max(total, na.rm = TRUE))) 

atrocities_top %>% 
  ggplot(aes(century, total)) +
  geom_point(aes(color = type)) +
  geom_text_repel(aes(label = atrocity), size = 2) +
  scale_y_log10() +
  scale_color_discrete(name = "Atrocity type") +
  labs(title = "Deaths from the worst atrocity by century",
       y = "Total deaths",
       x = "Century") +
  theme_minimal()

atrocities_tidy %>% 
  filter(century > 1100) %>% 
  ggplot(aes(reorder(atrocity, deaths), deaths, fill = type)) +
  geom_col(position = "dodge") +
  facet_grid(century ~., scales = "free_y", space = "free_y") +
  scale_y_log10() +
  theme_minimal() +
  theme(axis.text.y = element_text(size = 6),
        strip.text = element_text(size = 8),
        panel.grid.major.y = element_blank()) +
  scale_fill_discrete(name = "Atrocity Type") +
  coord_flip() +
  labs(x = "Atrocity",
       y = "Deaths",
       title = "Deaths from atrocities by century, 1200-1900")

Population data

q3.1 There's a limit to the conclusions we can draw without population data. You can find population data on the population sheet. Read in the data from the population sheet and store in a tibble. You only need the columns for year and population.

population <- 
  gs_read(sheet, ws = 3 , range = cell_cols("F:K"), 
                  skip = 11, col_names = FALSE, verbose = FALSE) %>% 
  dplyr::select(year = X1,
         population = X6)

population %>% 
  ggplot(aes(year, population)) +
  geom_line() +
  labs(x = "Year",
       y = "Population") +
  theme_minimal()

q3.3 Now, join the population data with atrocities_tidy. Store the result in a new tibble. Think about how you want to deal with the differences in population across the time span of an atrocity.

#I realized that we don't actually need this population data, but if students want to either try to make the legends of the NY times plot or play around with the data, it could be useful
#I also think either some of the death estimates are off, or some of the population statistics are off
atrocities_pop <-
  atrocities_tidy %>% 
  left_join(population, by = c("start_year" = "year")) %>% 
  rename(population_start = population) %>% 
  left_join(population, by = c("end_year" = "year")) %>% 
  rename(population_end = population)

q3.4 To recreate the plot, we're going to need a variable total that represents the total number of deaths due to an atrocity, a variable percent_killed that represents the percent of the world's total population killed in the atrocity, a variable mean_year that represents the median year of the atrocity, and a variable percent_killed that represents the percentage of the population killed by the atrocity. Create these variables.

find_continent <- function(atrocity) {
  if (atrocity %in% europe) {return("europe")}
  if (atrocity %in% africa) {return("africa")}
  if (atrocity %in% asia) {return("asia")}
  if (atrocity %in% americas) {return("americas")}
  if (atrocity %in% multiple) {return("multiple")}
  else {return("NA")}
}

by_atrocity <-
  atrocities_pop %>% 
  group_by(atrocity, start_year, end_year, length, 
           population_start, population_end, type) %>%
  summarise(total_deaths = sum(deaths, na.rm = TRUE)) %>% 
  mutate(mean_year = mean(c(start_year, end_year), na.rm = TRUE),
         mean_population = mean(c(population_start, population_end), na.rm = TRUE),
         percent_killed = (total_deaths / mean_population)*100,
         continent = map_chr(atrocity, find_continent))

Deaths from atrocities over time

q3.4 Recreate the New York Times plot. This doesn't have to be exact--the original author didn't use R to create the visualization.

by_atrocity_rf <-
  by_atrocity %>% 
       mutate(death_millions = round(total_deaths / 1000000, 1),
       label = str_c(atrocity, death_millions, sep = " "),
       y_pos = (end_year %% 250)*7)

for_labels_right <- 
  by_atrocity_rf %>% 
  filter(mean_year >= 1900)

for_labels_left <-
  by_atrocity_rf %>% 
  filter(mean_year < 1900)
  

by_atrocity_rf %>% 
  ggplot(aes(mean_year, y_pos, 
             size = total_deaths, color = continent)) +
  geom_point(alpha = .8) +
  geom_text_repel(aes(end_year, y_pos, label = label), 
                  size = 3, segment.size = .5, color = "black", 
                  nudge_x = 1, nudge_y = -30, inherit.aes = FALSE, 
                  min.segment.length = unit(2, "lines"),
                  data = for_labels_right) +
  geom_text_repel(aes(start_year, y_pos, label = label), 
                  size = 3, segment.size = .5, color = "black",
                  nudge_x = -3, nudge_y = -30, inherit.aes = FALSE, 
                  min.segment.length = unit(2, "lines"),
                  data = for_labels_left) +
  geom_segment(aes(x = start_year, xend = end_year, 
                   y = y_pos, yend = y_pos), 
               size = .1, inherit.aes = FALSE) +
  geom_vline(xintercept = 0, linetype = "dotted") +
  geom_vline(xintercept = 1000, linetype = "dotted") +
  facet_wrap(~type, ncol = 1, scales = "free_y") +
  scale_x_continuous(breaks = seq(-400, 2600, 200), position = "top") +
  scale_y_continuous(trans = "reverse", labels = NULL) +
  scale_color_manual(values = fill_continent_colors) +
  scale_radius(range = c(6, 36)) +
  coord_cartesian(ylim = c(-100, 1900), xlim = c(-400, 2400)) +
  theme_minimal() +
  guides(size = FALSE,
         color = guide_legend(title = "Continent", override.aes = list(size = 8))) +
  theme_minimal() +
  theme(
   # aspect.ratio = .15,
    panel.grid.major.x = element_line(linetype = "dotted"),
    panel.grid.minor.x = element_line(linetype = "dotted"),
    panel.grid.major.y = element_blank(),
    panel.grid.minor.y = element_blank(),
    strip.text = element_text(size = 24, family = "Times", face = "bold", hjust = 0),
    axis.text.x = element_text(size = 10, face = "bold"),
    panel.background = element_rect(fill = NA, color = "black")
  ) +
  labs(x = NULL,
       y = NULL)

Large-scale atrocities

Read in data from googlesheets

Atrocities

Population data

Deaths from atrocities over time