Lines - tracking trends

Session 5 practical exercises

Barplots are useful for categorical counts and combinations of those. When we account for four years - that is: four data points, the resulting graph can be adequate. Increase that number of data points, and it becomes increasingly difficult to read as well, so let’s start moving towards finer time scales and learn about useful geoms to assist us in our tasks.

Kassandra’s next plot request is about case trends overall and by sex. She is interested in learning more about disease seasonality and asks us to deliver a monthly trend line with the three categories.

Continue in your Session 5 script, where your data and libraries are already loaded.

Final plot - Trend by sex

Finnish the plot:

Action — Add scale_x_date() with 4-months date_breaks and date_labels in month-year with "%b %Y".

Action — Add a theme of your choice, and labs() for the axes, legend title and a report-ready title.

Part 9 · From line to area

A line is a bar in disguise

Think back to geom_bar()/geom_col() versus geom_line(). Both take the same kind of input — one value per position — and simply draw it differently: as a bar, or as a connected point. geom_area() is a third way of drawing that same idea: a line, with the space underneath it filled in.

Recover the code of the final plot of serogroup cases by year. We will turn it into an area plot of monthly cases, using floor_date() again. Copy this code to see for yourself, but don’t spend any more time on this - you still have exercises to do!

# Monthly series by serogroup
imd_monthly_serogroup <- imd %>%
  mutate(
    month = floor_date(key_date, unit = "month")
  ) %>%
  count(month, serogroup)

# Final plot (2): stacked area by serogroup
serogroup_area <- imd_monthly_serogroup %>%
  ggplot(aes(
    x = month,
    y = n,
    fill = serogroup
  )) +
  geom_area() +
  scale_fill_brewer(
    palette = "Set2"
  ) +
  scale_x_date(
    date_breaks = "6 months",
    date_labels = "%b %Y"
  ) +
  labs(
    x = "Month",
    y = "Number of cases",
    fill = "Serogroup",
    title = "Monthly IMD cases by serogroup"
  ) +
  theme_classic()

serogroup_area

By default, geom_area() stacks — each serogroup’s monthly count sits on top of the previous one, so the height of the whole stack at any given month is the national total for that month. Instead of seven crossing lines, you get one continuous silhouette, split into colored bands. The best of the bar and the best of the line.

Exercise Summary

Two graphs, two different problems solved. The first taught you a genuinely reusable trick: turning a separate summary into a fabricated category of the same variable, so bind_rows() and a single aes() can put it on equal footing with the real categories and, from there, scale_color_manual() to give a meaningful category its own deliberate color.

The second showed that bar, line and area are three faces of the same idea, and that switching from line to area is often the fix when too many crossing lines stop being readable. In the next excercise, the calendar itself becomes the subject of the plot, one day at a time.

Function Package What it does
floor_date() lubridate Rounds a date down to the start of a given unit (day, month, year…)
bind_rows() dplyr Stacks two tables with matching columns into one
scale_color_manual() ggplot2 Assigns specific colors to specific category values, written by hand
geom_area() ggplot2 Like geom_line(), with the area underneath filled; stacks by default when fill is mapped 
# Load libraries
library(pacman)
p_load(rio, here, tidyverse)

# Import data
imd <- import(here("data", "clean", "IMD_Sample_Clean.rds"))

# The trap: daily resolution is too fine
imd %>%
  count(key_date) %>%
  ggplot(aes(
    x = key_date,
    y = n
  )) +
  geom_line()

# Monthly series by sex (unknown sex excluded)
imd_monthly_sex <- imd %>%
  filter(!is.na(sex)) %>%
  mutate(
    month = floor_date(key_date, unit = "month")
  ) %>%
  count(month, sex)

# Monthly total (same exclusion, so totals match Male + Female)
imd_monthly_total <- imd %>%
  filter(!is.na(sex)) %>%
  mutate(
    month = floor_date(key_date, unit = "month")
  ) %>%
  count(month) %>%
  mutate(
    sex = "Total"
  )

# The trick: stack both tables into one
imd_monthly_sex_total <- bind_rows(
  imd_monthly_sex,
  imd_monthly_total
)

# Final plot (1): trend by sex, with a manual color scale
trend_by_sex <- imd_monthly_sex_total %>%
  ggplot(aes(
    x = month,
    y = n,
    color = sex
  )) +
  geom_line(
    linewidth = 0.8
  ) +
  scale_color_manual(
    values = c(
      "Total"  = "grey20",
      "Male"   = "steelblue",
      "Female" = "darkorange"
    )
  ) +
  scale_x_date(
    date_breaks = "4 months",
    date_labels = "%b %Y"
  ) +
  labs(
    x = "Month",
    y = "Number of cases",
    color = "Sex",
    title = "Monthly IMD cases, overall and by sex"
  ) +
  theme_classic()

# Save the plot
ggsave(
  plot     = trend_by_sex,
  filename = "monthly_trend_sex.png",
  path     = here("output"),
  units    = "in",
  width    = 7,
  height   = 5,
  dpi      = 300
)

# Monthly series by serogroup
imd_monthly_serogroup <- imd %>%
  mutate(
    month = floor_date(key_date, unit = "month")
  ) %>%
  count(month, serogroup)

# Final plot (2): stacked area by serogroup
serogroup_area <- imd_monthly_serogroup %>%
  ggplot(aes(
    x = month,
    y = n,
    fill = serogroup
  )) +
  geom_area() +
  scale_fill_brewer(
    palette = "Set2"
  ) +
  scale_x_date(
    date_breaks = "6 months",
    date_labels = "%b %Y"
  ) +
  labs(
    x = "Month",
    y = "Number of cases",
    fill = "Serogroup",
    title = "Monthly IMD cases by serogroup"
  ) +
  theme_classic()