knitr::opts_chunk$set(
  warning = FALSE,
  message = FALSE,
  error = FALSE,
  dev = "svg"
  )

The govstyle package is designed to give your ggplot2 figures gov.uk friendly formatting. At present the package consists of a single theme which can be applied to ggplots. More functionality will be added in the future.

A real life example

In this vignette, we will reproduce two of the plots presented in the 2015 Statistical First Release (SFR) 39 from the Department of Education. This SFR deals with statistics relating to student absence and exclusion, and is available for download here.

Getting the data

The first step is to download and prepare the data. The SFR data are stored as a large (41.7 MB) CSV file within a zip file, which is available here.

Downloading and extracting the data can all be done in R

download.file(
"https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/468966/SFR39_2015_Underlying_data.zip",
"UD.zip",
quiet = FALSE,
mode = "w",
cacheOK = TRUE
)

unzip(
  "UD.zip",
  files = "SFR39_2015_Autumn_Spring_Proposed_SFR_structure.csv",
  overwrite = TRUE
  )

It's worth doing a quick check to ensure that this worked:

file.exists("SFR39_2015_Autumn_Spring_Proposed_SFR_structure.csv")

## Should be 43776735 bytes:

file.info("SFR39_2015_Autumn_Spring_Proposed_SFR_structure.csv")$size

We also need to install the govstyle package. This vignette uses version v0.1.0 - leaving out this argument from devtools::install_github will fetch the latest commit on the master branch

devtools::install_github(
  repo = "ukgovdatascience/govstyle"
)
library(govstyle)

Loading and cleaning the data

If all has gone well, we can load the data and make some basic maniuplations. For this we need both readr and dplyr:

library(readr)
library(dplyr)
library(tidyr)
library(readr)
library(dplyr)
library(tidyr)

Here I use the dplyr framework using the pipe %>% to combine a lot of cleaning tasks into a single block of code. First I load the data from CSV using readr::read_csv(), then use dplyr::select() to subset the columns I am interested in and dplyr::mutate() to standardise the school types in lowercase.

absence_data_full <- read_csv(
  file = "SFR39_2015_Autumn_Spring_Proposed_SFR_structure.csv",
  na = c( "x", ".", ""),
  col_types = cols(sess_auth_ext_holiday = col_integer())
)

# For brevity of printing, select only columns of interest.

absence_data <- absence_data_full %>%
  select(
    Period, Level, Year, Country,
    School_type, sess_possible, sess_overall
  ) %>%
  mutate(
    School_type = tolower(School_type)
  )

From a quick scan of the data, we can see that all the remaining character columns have been converted to factor, and we have two remaining numeric columns which are integer.

absence_data

Calculating Overall Absence Rate

To recreate the plots in the SFR, we first need to calculate the national overall absence rate (OAR), which is given as:

the total number of overall absence sessions for all pupils as a percentage of the total number of possible sessions for all pupils, where overall absence is the sum of authorised and unauthorised absence and one session is equal to half a day.

or:

$$ \text{Overall absence rate} = \frac{\text{Total Overall absence sessions}}{\text{Total sessions possible}}\times 100 $$

For this we need to subset the data and calculate this at the national level, and combine this calculation with the regional data

# Calculate the national OAR values.

oar_summary <- absence_data %>%
  dplyr::filter(
    Level == "NATIONAL"
  ) %>%
  mutate(
    oar = (sess_overall/sess_possible) * 100
  )

# Calculate the OAR values for Period, Level, Year, and Country combinations

oar_summary_combined <- absence_data %>%
  dplyr::filter(
    Level == "NATIONAL"
  ) %>%
  group_by(Period, Level, Year, Country) %>%
  summarise(
    sess_possible = sum(sess_possible),
    sess_overall = sum(sess_overall)
  ) %>%
  mutate(
    oar = (sess_overall/sess_possible) * 100,
    School_type = "state-funded primary and secondary"
  )

# Combine the two above dataframes

oar_summary <- bind_rows(
  oar_summary,
  oar_summary_combined
)

Note that combining the two dataframes above leads to School_type being coerced to character. This is not an issue for use here.

oar_summary

Creating the first plot

For the first plot, we also need the values from the start and end of the timeseries for inclusion in the plot.

oar_values <- oar_summary %>%
  filter(
    Year %in% c("2006/07","2014/15")
  ) %>%
  arrange(Year)


oar_values

To produce a nice plot ends up in quite a lot of code, so I will build up bit by bit.

library(ggplot2)

p <- oar_summary %>%
  ggplot +
  aes(
    x = Year,
    y = oar,
    colour = School_type,
    fill = School_type,
    group = School_type
  ) +
  geom_path(size = 1.5) +
  xlab("Autumn and Spring term") +
  ylab("Overall absence rate (%)")

This gives us our base plot

p

Government tends to like seeing zero on the y-axis, so lets fix the axes with expand_limits(), and add a title with ggtitle.

p1 <- p +
  expand_limits(
    x = 0,
    y = c(0, 8.5)
    )   +
  ggtitle(
    "Overall absence rate across state-funded\nprimary and secondary schools"
  )

p1

At this point I apply theme_gov(), and introduce a scale using colours from the gov.uk colour palette. For this we can call check_pal()

check_pal()
p2 <- p1 +
  theme_gov(
    base_size = 12,
    base_colour = "gray40") +
  scale_colour_manual(
    values =  gov_cols[c("turquoise","brown","light_blue")] %>% unname
  )

p2

theme_gov() removes the legend by default, so I'll label the lines instead. This gets a little complicated here as we need to nudge the values into the correct place using the hjust and vjust arguments. I also use the sprintf() command to force R to print a single decimal place, even if this number is zero - the default would be not to do this.

p3 <- p2 +
  geom_text(
    data = oar_values,
    aes(
      label = sprintf("%.1f", oar)
      ),
    hjust = rep(c(1.35,-0.35), each = 3),
    fontface = "bold"
  )+
  geom_text(
    data = oar_summary %>% filter(Year == "2006/07"),
    aes(
      label = c(
        "Primary",
        "Secondary",
        "Primary and secondary"
      )
    ),
    hjust = 0,
    vjust = -1,
    fontface = "bold"
  )

p3

So this is pretty close to the final figure. One thing we might want to do is rotate the y-axis label so that it reads horizontally

p4 <- p3 +
  theme(
    # plot.margin = grid::unit(
    #   c(0, 5, 5, 0), "mm"),
    axis.title.y = element_text(
      angle = 0
      )
  )

p4

Creating the second plot

Illness absence rates

Start with the full absence data. Filter to only NATIONAL values, then sum over years for the variables sess_overall, sess_possible, and sess_auth_illness. Then calculate the overall absence rate, and the illness absence rate, and finally gather this up into a long rather than a wide data.frame to allow easier plotting of colours

illness_summary <- absence_data_full %>%
  dplyr::filter(Level == "NATIONAL") %>%
  group_by(Year) %>%
  summarise(
    sess_overall = sum(sess_overall),
    sess_possible = sum(sess_possible),
    sess_auth_illness = sum(sess_auth_illness)
  ) %>%
  mutate(
    oar = (sess_overall / sess_possible) * 100,
    iar = (sess_auth_illness / sess_possible) * 100
  ) %>%
  gather(key, value, oar:iar)

illness_summary

Now for the plotting. Rather than approach it piece by piece, I include the full code here in a single chunk.

# Start with the new illness_summary object

illness_summary %>%

  # Set up the basics of the plot

  ggplot +
  aes(
    x = Year,
    y = value,
    group = key,
    colour = key
  ) +

  # Add the lines

  geom_path(size = 1.5) +

  # Add the values at the start and end of the lines

  geom_text(
    data = illness_summary %>% filter(Year %in% c("2006/07","2014/15")) %>% arrange(Year),

    # Force values to show one decimal place even if that is zero

    aes(label = sprintf("%.1f", value)),

    # Nudge the values away from the lines

    hjust = rep(c(1.25,-0.25),each = 2),
    fontface = "bold"
  ) +

  # Label the lines

  geom_text(
    data = illness_summary %>% filter(Year == "2006/07"),
    aes(label = c(
      "Overall absence rate",
      "Illness absence rate"
    )),

    # Left justify, and nudge the values up away from the lines

    hjust = 0,
    vjust = -1.2,
    size = 4,
    fontface = "bold"
  ) +

  # axis limits

  expand_limits(x = 0, y = c(0, 8)) +

  # Use the gov.uk colours

  scale_colour_manual(values = gov_cols[c("turquoise","brown")] %>% unname) +

  # Apply theme_gov

  theme_gov(
    base_size = 12, base_colour = "gray40", axes = "x"
  ) +

  # Label the axes

  xlab("Autumn and spring term") +
  ylab("Absence rate (%)") +

  # Add a title. Note that line breaks in the title must be specified manually
  # with "\n"

  ggtitle(
    "Comparison of the trend in overall and illness\n absence rates: England, autumn 2006 and\n spring 2007 to autumn 2014 and spring 2015"
  ) +

  # Make the y-axis title horizontal, and at the top of the axis.
  # Adjust margins to compensate for this.
  # Adjust the axis breakpoints.

  theme(
    axis.title.y = element_text(
      angle = 0, hjust = 20, vjust = 1.01
    ),
    plot.margin = grid::unit(c(0,5,5,0), "mm")
  ) +
  scale_y_continuous(breaks = c(0, seq(0, 8, 2)))
if (file.exists("UD.zip")) file.remove("UD.zip")


ukgovdatascience/govstyle documentation built on Oct. 8, 2019, 7:39 p.m.