How to create segmented graphs in ggplot2 with legend?

Question

I have a data as follows:

I would like to create a segmented plot (like a pre- and post- plot, including the vertical line at t = 10, to indicate the change. t refers to the elapsed time, x refers to 0 for pre-implementation, 1 for post-implementation and count_visit_triage\\d are count data that I would like to plot in the y-axis.

This is my r-code. I have pieced together multiple geom_smooth into the same figure, each colour representing values from triage1, triage2 etc. Because of this, I couldn't obtain the legend. My question is (1) how can we simplify this code so that the legend can be included in the figure?

ggplot(df, aes(x = t, y = count_visit_triage1)) +
  geom_smooth(data = subset(df, x == 0), aes(x = t, y = count_visit_triage1), colour = "blue", se = F) +
  geom_smooth(data = subset(df, x == 1), aes(x = t, y = count_visit_triage1), colour = "blue", se = F) +
  geom_smooth(data = subset(df, x == 0), aes(x = t, y = count_visit_triage2), colour = "orange", se = F) +
  geom_smooth(data = subset(df, x == 1), aes(x = t, y = count_visit_triage2), colour = "orange", se = F) +
  geom_smooth(data = subset(df, x == 0), aes(x = t, y = count_visit_triage3), colour = "green", se = F) +
  geom_smooth(data = subset(df, x == 1), aes(x = t, y = count_visit_triage3), colour = "green", se = F) +
  geom_smooth(data = subset(df, x == 0), aes(x = t, y = count_visit_triage4), colour = "red", se = F) +
  geom_smooth(data = subset(df, x == 1), aes(x = t, y = count_visit_triage4), colour = "red", se = F) +
  geom_vline(xintercept = 10, linetype = "dashed") + 
  theme_bw()

Data:

df <- structure(list(t = 1:20, x = c(0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 
0L, 0L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), count_visit_triage1 = c(42L, 
55L, 61L, 52L, 58L, 38L, 47L, 46L, 66L, 44L, 24L, 17L, 40L, 25L, 
18L, 23L, 34L, 35L, 22L, 23L), count_visit_triage2 = c(175L, 
241L, 196L, 213L, 189L, 163L, 181L, 166L, 229L, 224L, 153L, 139L, 
125L, 145L, 134L, 115L, 152L, 153L, 136L, 154L), count_visit_triage3 = c(120L, 
114L, 106L, 88L, 108L, 103L, 103L, 93L, 80L, 81L, 88L, 94L, 94L, 
77L, 91L, 100L, 93L, 70L, 79L, 77L), count_visit_triage4 = c(3L, 
0L, 0L, 1L, 2L, 2L, 0L, 4L, 4L, 2L, 0L, 0L, 0L, 0L, 0L, 1L, 0L, 
0L, 1L, 2L)), row.names = c(NA, -20L), class = c("tbl_df", "tbl", 
"data.frame"))

score 2 · Accepted Answer · answered Oct 28 '20 at 05:00

2

Reshape the data then specify the col and group aesthetics.

library(tidyverse)

df %>%
  pivot_longer(starts_with("count_")) %>%
  ggplot(aes(t, value, col = name, group = paste(x, name))) +
    geom_smooth(se = FALSE) +
    geom_vline(xintercept = 10, linetype = "dashed") + 
    theme_bw()

answered Oct 28 '20 at 05:00

Paul

8,734
1
26
36

Thanks Paul, what exactly is the `group` aesthetics? – HNSKD Oct 28 '20 at 05:07
The `group` in `aes` (aesthetic mapping) determines what points `geom_smooth` should group together into a single line. In this code, points with the same `x` and `name` will be grouped together into one line. – Paul Oct 28 '20 at 05:08

score 2 · Answer 2 · answered Oct 28 '20 at 05:03

You can try this:

library(tidyverse)
df %>% 
  pivot_longer(cols = -c(t,x),
               names_to = "visit",
               values_to = "count") %>%
  ggplot() +
  geom_line(aes(x = t, 
                y = count, 
                color = visit,
                group = interaction(x,visit))) +
  geom_vline(xintercept = 10, linetype = "dashed") + 
  scale_color_manual(name = "legend",
                     values = 1:4,
                     labels = c("Visit Triage 1",
                                "Visit Triage 2",
                                "Visit Triage 3",
                                "Visit Triage 4")) +
  theme_bw()

Yes. `interaction` is the proper way to combine two strings into a factor, rather than `paste` like I did. — Paul, Oct 28 '20 at 05:11

How to create segmented graphs in ggplot2 with legend?

2 Answers2