Skip to main content

8.3 How Do Absences Spread Across the Week?

The next obvious slice is by day of the week. Are Mondays really the worst day for absences, or is that just a stereotype? The dataset codes the weekday as a number (Day of the week is 2 for Monday through 6 for Friday — no weekend rows because the workplace is closed), which we can recode the same way we did for months.

weekday_labels <- c("Mon", "Tue", "Wed", "Thu", "Fri")

absences_by_weekday <- work |>
  filter(`Month of absence` > 0) |>
  group_by(`Day of the week`) |>
  summarise(
    n_absences = n(),
    total_hours = sum(`Absenteeism time in hours`)
  ) |>
  mutate(
    Weekday = factor(weekday_labels[`Day of the week` - 1],
                     levels = weekday_labels)
  ) |>
  select(Weekday, n_absences, total_hours)

A small mechanical detail: because the day codes start at 2, we subtract 1 before indexing into weekday_labels. Setting levels = weekday_labels again forces calendar order rather than alphabetical order on the chart.

Table 8.3: Table 8.4: Absences by day of the week.

Weekday	Number of absences	Total hours absent
Mon	161	1489
Tue	153	1229
Wed	155	1115
Thu	125	553
Fri	143	738

We can use exactly the same plotting recipe as before — pivot_longer(), then a faceted bar chart with free y-scales:

absences_by_weekday |>
  pivot_longer(
    cols = c(n_absences, total_hours),
    names_to = "metric",
    values_to = "value"
  ) |>
  ggplot(aes(x = Weekday, y = value, fill = metric)) +
  geom_col() +
  facet_wrap(~ metric, ncol = 1, scales = "free_y",
             labeller = as_labeller(c(
               n_absences = "Number of absences",
               total_hours = "Total hours absent"
             ))) +
  labs(x = NULL, y = NULL) +
  theme_minimal() +
  theme(legend.position = "none")

8.3.1 What the chart tells us

The “Monday is worst” stereotype gets partial support — Monday does have the highest count of absence events. But the total hours picture is different: Wednesday loses the most hours, despite having fewer events than Monday or Tuesday. That contrast — many short Monday absences, fewer-but-longer Wednesday absences — is the same story we saw at the monthly level (March vs. July) showing up at a different time scale.

Two things to notice about the method, not just the data:

1. The plotting recipe didn’t change. The same group_by → summarise → pivot_longer → facet_wrap pattern handles “by month” and “by day of week” with only the grouping variable swapped. This is a hallmark of well-designed data tools: once you have a recipe that works, applying it to a new question is mostly mechanical.
2. The chart shape changed in a useful way. Five weekday bars are easier to read at a glance than twelve monthly bars. When you have a categorical variable with only a handful of levels, a faceted bar chart is hard to beat.