Initial Visualisations

Experiment!

library(tidyverse)
worker_data <- read_csv("http://linux.oii.ox.ac.uk/~otto.kassi/OLI/worker_countrydata.txt")

Show number of workers today:

worker_data %>%
  filter(timestamp == max(timestamp)) %>%
  group_by(country) %>%
  mutate(workers.in.country = sum(num_workers)) %>%
  select(country, workers.in.country) %>%
  unique() %>%
  ungroup() %>%
  arrange(workers.in.country) %>%
  mutate(country = factor(country, levels = unique(country))) %>%
  ggplot(aes(x = country, y = workers.in.country)) + geom_col() + coord_flip() + ggtitle("Most Recent Number of Workers per country")

What is today’s success rate?

library(plotly)
g_data <- worker_data %>%
  filter(timestamp == max(timestamp)) %>%
  mutate(success.rate = num_projects / num_workers) %>%
  group_by(country) %>%
  mutate(country.avg.success.rate = 100 * mean(success.rate)) %>%
  select(country, country.avg.success.rate) %>%
  ungroup() %>%
  unique() %>%
  arrange(country.avg.success.rate) %>%
  mutate(country = factor(country, levels = unique(country)))

ggplot(g_data, aes(x = country, y = country.avg.success.rate)) +
  geom_bar(stat = "identity") +
  coord_flip() +
  ggtitle("Today's Country Success Rate",
          subtitle = "(including 100% successful countries)")

Plotly version of above

plot_ly(g_data,
        x = ~country.avg.success.rate,
        y = ~country) %>%
  layout(title = "Today's Country Success Rate (including 100% successful countries)")

g_data <- worker_data %>%
  filter(timestamp == max(timestamp)) %>%
  mutate(success.rate = 100 * {num_projects / num_workers}) %>%
  filter(success.rate < 100) %>%
  group_by(country) %>%
  mutate(country.avg.success.rate = mean(success.rate)) %>%
  select(country, country.avg.success.rate) %>%
  ungroup() %>%
  unique() %>%
  arrange(country.avg.success.rate) %>%
  mutate(country = factor(country, levels = unique(country)))

ggplot(g_data, aes(x = country, y = country.avg.success.rate)) + 
  geom_bar(stat = "identity") + coord_flip() + ggtitle("Today's Country Success Rate",
                                      subtitle = "(excluding 100% successful countries)")

Plotly version of above

plot_ly(g_data,
        x = ~country.avg.success.rate,
        y = ~country) %>%
  layout(title = "Today's Country Success Rate (excluding 100% successful countries)")

Success Rates

What is the average success rate for a country today? The average success rate is 12.9288939

country_avg_succes_rate <- worker_data %>%
  filter(timestamp == max(timestamp)) %>%
  mutate(success.rate = 100* {num_projects / num_workers}) %>%
  filter(success.rate < 100) %>%
  select(success.rate) %>%
  .[[1]] %>%
  mean()

g_data <- worker_data %>%
  filter(timestamp == max(timestamp)) %>%
  mutate(success.rate = 100* {num_projects / num_workers}) %>%
  filter(success.rate < 100) %>%
  select(success.rate) %>%
  mutate(country.average.success = factor("Country Average Success"))

ggplot(g_data, aes(x = country.average.success, y = success.rate)) + 
  geom_violin() + ggtitle("Today's Country Success Rate") + labs(y = "Percent", x = "")

However, when interpreting this chart it’s important to note that many countries only have a small number of workers/won projects and this highly skews the data. For instance, Yemen:

worker_data %>%
  filter(timestamp == max(timestamp)) %>%
  filter(country == "Yemen")

## # A tibble: 1 × 5
##    timestamp country              occupation num_workers num_projects
##       <date>   <chr>                   <chr>       <int>        <int>
## 1 2017-03-09   Yemen Creative and multimedia           4            3

We could look at the distribution of workers across countries, and discount values outside of the 2nd and 3rd quartiles.

g_quantiles <- quantile(worker_data$num_workers)

g_data <- worker_data %>%
  filter(timestamp == max(timestamp)) %>%
  filter(num_workers >= g_quantiles[["25%"]]) %>%
  filter(num_workers <= g_quantiles[["75%"]]) %>%
  select(num_workers) %>%
  mutate(workers = "workers")

ggplot(g_data, aes(x = workers, y = num_workers)) + geom_violin() + ggtitle("Distribution of number of workers",
                                                                            subtitle = "For the 2nd - 3rd quantiles")

Plot the above data as a deviation chart:

g_quantiles <- quantile(worker_data$num_workers)
g_data <- worker_data %>%
  filter(timestamp == max(timestamp)) %>%
  filter(num_workers >= g_quantiles[["25%"]]) %>%
  filter(num_workers <= g_quantiles[["75%"]]) %>%
  mutate(success.rate = 100 * {num_projects / num_workers}) %>%
  filter(success.rate < 100) %>%
  group_by(country) %>%
  mutate(country.avg.success.rate = mean(success.rate)) %>%
  select(country, country.avg.success.rate) %>%
  ungroup() %>%
  unique() %>%
  arrange(country.avg.success.rate) %>%
  mutate(country = factor(country, levels = unique(country)))

country_avg_succes_rate <- worker_data %>%
  filter(timestamp == max(timestamp)) %>%
  mutate(success.rate = 100* {num_projects / num_workers}) %>%
  filter(success.rate < 100) %>%
  select(success.rate) %>%
  .[[1]] %>%
  mean()

g_quantiles <- quantile(g_data$country.avg.success.rate)
ggplot(g_data, aes(x = country, y = country.avg.success.rate)) + 
  geom_point() + 
  coord_flip() + 
  geom_hline(yintercept = g_quantiles[["50%"]]) + 
  geom_segment(aes(y = country.avg.success.rate, yend = g_quantiles[["50%"]], xend = country), size = 2,
               alpha = 0.2) +
  annotate("rect", xmin = -Inf, xmax = Inf, ymin = g_quantiles[["25%"]], ymax = g_quantiles[["75%"]], fill = "blue", alpha = .1, color = NA) + labs(x = "", y = "Success Rate %")

Success rate by occupation

# g_quantiles <- quantile(worker_data$num_workers)
g_data <- worker_data %>%
  filter(timestamp == max(timestamp)) %>%
  # filter(num_workers >= g_quantiles[["25%"]]) %>%
  # filter(num_workers <= g_quantiles[["75%"]]) %>%
  mutate(success.rate = 100 * {num_projects / num_workers}) %>%
  filter(success.rate < 100) %>%
  group_by(occupation) %>%
  mutate(occupation.avg.success.rate = mean(success.rate)) %>%
  select(occupation, occupation.avg.success.rate) %>%
  ungroup() %>%
  unique() %>%
  arrange(occupation.avg.success.rate) %>%
  mutate(occupation = factor(occupation, levels = unique(occupation)))

occupation_avg_succes_rate <- worker_data %>%
  filter(timestamp == max(timestamp)) %>%
  mutate(success.rate = 100* {num_projects / num_workers}) %>%
  filter(success.rate < 100) %>%
  select(success.rate) %>%
  .[[1]] %>%
  mean()

g_quantiles <- quantile(g_data$occupation.avg.success.rate)
ggplot(g_data, aes(x = occupation, y = occupation.avg.success.rate)) + 
  geom_point() + 
  coord_flip() + 
  geom_hline(yintercept = g_quantiles[["50%"]]) + 
  geom_segment(aes(y = occupation.avg.success.rate, yend = g_quantiles[["50%"]], xend = occupation), size = 2,
               alpha = 0.2)