1.1 Import the Data

Let’s fetch the date from the official page of the Berlin Senate. You’ll get a PDF which you have to destill with the tabulizer package (or by hand) in order to get a CSV. I will follow up with a post on using tabulizer from within RStudio anytime soon.

library(tidyverse)
# Datenquelle: https://www.berlin.de/sen/justiz/juristenausbildung/juristische-pruefungen/artikel.264039.php
data_path <- here::here("static", "data", "/")
noten_raw <- read_csv2(str_c(data_path, "noten_201704.csv")) # nach PDF -> Tabulizer
head(noten_raw) %>% knitr::kable("html", 2)

1.2 Skim / Preview the Data

The skimr-Package - among others - is great for quickly inspecting what kind of data (variables, data type, NAs etc.) you get.

skimr::skim_to_wide(noten_raw) %>% knitr::kable("html", 2)

## Warning: 'skimr::skim_to_wide' is deprecated.
## Use 'skim()' instead.
## See help("Deprecated")

(In RStudio / R Markdown the hist column is rendered properly. You get a nice histogram per (numeric) variable. There seems to be an issue with Knitr & UTF-8 encoding on MS Windows systems.)

(Screenshot of skimr from RStudio)

1.3 Long -> Short with gather()

Now we need to tidy the data. We first drop the AZ column (Student id) and then “pivot” all the exam subjects into a single column (= variable) named “Fach” (GER for subject).

median_2017 <- median(noten_raw$Dur, na.rm = TRUE)
noten_raw %>%
  select(-AZ) %>% 
  gather(key = Fach, value = Punkte) -> noten_long

head(noten_long) %>% knitr::kable("html", 2)

1.4 2017/IV Joyplot

library(ggridges)
library(viridis)

# Beschriftungen
title_a <- c("2. Juristisches Staatsexamen, GJPA Berlin/Brandenburg")
subtitle_a = paste0("Notenverteilung Kampagne 4/17; n = ",nrow(noten_raw),
                    "; \"Dur\" = durchschnittl. Examensnote\r\nLinien: rot  = \"bestanden (ausreichend)\", blau  = \"vollbefriedigend\",\r\nschwarz  = Median Gesamtnote (",median_2017,")")
caption_a = c("@fubits; Daten: GJPA 2018")

# Plot
noten_long %>% 
  ggplot() +
  geom_density_ridges(aes(x = Punkte, y = Fach, fill = Fach),
                      rel_min_height = 0.025,
                      scale = 1.75) +
  # Linie: Vollbefriedigend
  geom_vline(xintercept = 10, color = "blue", linetype = 4, size = 1) + 
  # Linie: Bestanden
  geom_vline(xintercept = 4, color = "red", linetype = 4, size = 1) +
  # Linie: Median Gesamtnote
  geom_vline(xintercept = median_2017, color = "black", size = 1) +
  labs(title = title_a, subtitle = subtitle_a, caption = caption_a) +
  scale_x_continuous(breaks = c(0:18), limits = c(0,18)) +
  scale_y_discrete(expand = c(0.01,0.0)) +
  scale_fill_viridis(option = "D", name = "Frequency n",
                     direction = -1, discrete = TRUE) +
  # theme(legend.position = "none")
  theme_minimal() +
  guides(fill = FALSE)

1.5 2017/IV Boxplot

(Dur = overall result / final grade)

noten_long %>% 
ggplot() +
  geom_boxplot(aes(x = Fach, y = Punkte, fill = Fach)) +
  scale_y_continuous(breaks = c(0:18), limits = c(0, 18)) +
  scale_fill_viridis(
    option = "C",
    direction = -1, discrete = TRUE
  ) +
  labs(title = title_a, subtitle = subtitle_a, caption = caption_a) +
  # theme(legend.position = "none")
  theme_minimal() +
  guides(fill = FALSE) +
  # Linien zur Orientierung
  geom_hline(yintercept = 10, color = "blue", linetype = 4, size = 1) +
  geom_hline(yintercept = 4, color = "red", linetype = 4, size = 1) +
  geom_hline(yintercept = median_2017, color = "black", size = 1)

2.1 2018/I Joyplot

(Dur = overall result / final grade)

noten_long_2018 %>% 
  ggplot() +
  geom_density_ridges(aes(x = Punkte, y = Fach, fill = Fach),
                      rel_min_height = 0.025,
                      scale = 1.75) +
  # Linie: Vollbefriedigend
  geom_vline(xintercept = 10, color = "blue", linetype = 4, size = 1) + 
  # Linie: Bestanden
  geom_vline(xintercept = 4, color = "red", linetype = 4, size = 1) +
  # Linie: Median Gesamtnote
  geom_vline(xintercept = median_2018, color = "black", size = 1) +
  labs(title = title_a, subtitle = subtitle_a, caption = caption_a) +
  scale_x_continuous(breaks = c(0:18), limits = c(0,18)) +
  scale_y_discrete(expand = c(0.01,0.0)) +
  scale_fill_viridis(option = "D", name = "Frequency n",
                     direction = -1, discrete = TRUE) +
  # theme(legend.position = "none")
  theme_minimal() +
  guides(fill = FALSE)

2.2 2018/I Boxplot

noten_long_2018 %>% 
ggplot() +
  geom_boxplot(aes(x = Fach, y = Punkte, fill = Fach)) +
  scale_y_continuous(breaks = c(0:18), limits = c(0, 18)) +
  scale_fill_viridis(
    option = "C",
    direction = -1, discrete = TRUE
  ) +
  labs(title = title_a, subtitle = subtitle_a, caption = caption_a) +
  # theme(legend.position = "none")
  theme_minimal() +
  guides(fill = FALSE) +
  # Linien zur Orientierung
  geom_hline(yintercept = 10, color = "blue", linetype = 4, size = 1) +
  geom_hline(yintercept = 4, color = "red", linetype = 4, size = 1) +
  geom_hline(yintercept = median_2018, color = "black", size = 1)