Plot labels at ends of lines

Question

I have the following data (temp.dat see end note for full data)

   Year State     Capex
1  2003   VIC  5.356415
2  2004   VIC  5.765232
3  2005   VIC  5.247276
4  2006   VIC  5.579882
5  2007   VIC  5.142464
...

and I can produce the following chart:

ggplot(temp.dat) + 
  geom_line(aes(x = Year, y = Capex, group = State, colour = State))

Instead of the legend, I'd like the labels to be

1. coloured the same as the series
2. to the right of the last data point for each series

I've noticed baptiste's comments in the answer in the following link, but when I try to adapt his code (geom_text(aes(label = State, colour = State, x = Inf, y = Capex), hjust = -1)) the text does not appear.

ggplot2 - annotate outside of plot

temp.dat <- structure(list(Year = c("2003", "2004", "2005", "2006", "2007", 
"2008", "2009", "2010", "2011", "2012", "2013", "2014", "2003", 
"2004", "2005", "2006", "2007", "2008", "2009", "2010", "2011", 
"2012", "2013", "2014", "2003", "2004", "2005", "2006", "2007", 
"2008", "2009", "2010", "2011", "2012", "2013", "2014", "2003", 
"2004", "2005", "2006", "2007", "2008", "2009", "2010", "2011", 
"2012", "2013", "2014"), State = structure(c(1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 
4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L), .Label = c("VIC", 
"NSW", "QLD", "WA"), class = "factor"), Capex = c(5.35641472365348, 
5.76523240652641, 5.24727577535625, 5.57988239709746, 5.14246402568366, 
4.96786288162828, 5.493190785287, 6.08500616799372, 6.5092228474591, 
7.03813541623157, 8.34736513875897, 9.04992300432169, 7.15830329914056, 
7.21247045701994, 7.81373928617117, 7.76610217197542, 7.9744994967006, 
7.93734452080786, 8.29289899132255, 7.85222269563982, 8.12683746325074, 
8.61903784301649, 9.7904327253813, 9.75021175267288, 8.2950673974226, 
6.6272705639724, 6.50170524635367, 6.15609626379471, 6.43799637295979, 
6.9869551384028, 8.36305663640294, 8.31382617231745, 8.65409824343971, 
9.70529678167458, 11.3102788081848, 11.8696420977237, 6.77937303542605, 
5.51242844820827, 5.35789621712839, 4.38699327451101, 4.4925792218211, 
4.29934654081527, 4.54639175257732, 4.70040615159951, 5.04056109514957, 
5.49921208937735, 5.96590909090909, 6.18700407463007)), class = "data.frame", row.names = c(NA, 
-48L), .Names = c("Year", "State", "Capex"))

Liles

I would just create a separate data frame with only the data you want to plot like so geom_text(data = temp.dat[cumsum(table(temp.dat$State)), ], aes(label = State, colour = State, x = Year, y = Capex)) but there may be a more gg-way to do things – Criminology

Answer 1

To use Baptiste's idea, you need to turn off clipping. But when you do, you get garbage. In addition, you need to suppress the legend, and, for geom_text, select Capex for 2014, and increase the margin to give room for the labels. (Or you can adjust the hjust parameter to move the labels inside the plot panel.) Something like this:

library(ggplot2)
library(grid)

p = ggplot(temp.dat) + 
  geom_line(aes(x = Year, y = Capex, group = State, colour = State)) + 
  geom_text(data = subset(temp.dat, Year == "2014"), aes(label = State, colour = State, x = Inf, y = Capex), hjust = -.1) +
  scale_colour_discrete(guide = 'none')  +    
  theme(plot.margin = unit(c(1,3,1,1), "lines")) 

# Code to turn off clipping
gt <- ggplotGrob(p)
gt$layout$clip[gt$layout$name == "panel"] <- "off"
grid.draw(gt)

But, this is the sort of plot that is perfect for directlabels.

library(ggplot2)
library(directlabels)

ggplot(temp.dat, aes(x = Year, y = Capex, group = State, colour = State)) + 
  geom_line() +
  scale_colour_discrete(guide = 'none') +
  scale_x_discrete(expand=c(0, 1)) +
  geom_dl(aes(label = State), method = list(dl.combine("first.points", "last.points")), cex = 0.8) 

Edit To increase the space between the end point and the labels:

ggplot(temp.dat, aes(x = Year, y = Capex, group = State, colour = State)) + 
  geom_line() +
  scale_colour_discrete(guide = 'none') +
  scale_x_discrete(expand=c(0, 1)) +
  geom_dl(aes(label = State), method = list(dl.trans(x = x + 0.2), "last.points", cex = 0.8)) +
  geom_dl(aes(label = State), method = list(dl.trans(x = x - 0.2), "first.points", cex = 0.8)) 

Valise

Did not know about the directlabels package. I couldn't see in the documentation a way to manually increase the horizontal space between the end points and the text label. What's the best way to do this? – Liles

I've added an edit. See FAQs (number 5) at http://directlabels.r-forge.r-project.org/ – Valise

Trying to install the package: package ‘directlabels’ is not available (for R version 3.3.2). I can't find the FAQ site for the package as well. Is it still live? – Glory

@Glory Hmm. I'm not sure what's happening. The link is still live. "Frequently asked questions" is on the first page. And I've just checked with cran - directlabels is available. – Valise

@slhck, Looks like it hasn't been installed. Have you tried installing quadprog? – Valise

@slhck I've just checked CRAN, quadprog v1.5-7 is there. (The latest version for R is 3.6.0.) But quadprog should install with directlabels. Have you tried reinstalling directlabels? – Valise

It works now. Perhaps the version was just missing at the time when I tested. Thanks! – Mayor

clipping can now be easily turned off with coord_...(clip = "off") – Chaddie

Answer 2

A newer solution is to use ggrepel:

library(ggplot2)
library(ggrepel)
library(dplyr)

temp.dat %>%
  mutate(label = if_else(Year == max(Year), as.character(State), NA_character_)) %>%
  ggplot(aes(x = Year, y = Capex, group = State, colour = State)) + 
  geom_line() + 
  geom_label_repel(aes(label = label),
                  nudge_x = 1,
                  na.rm = TRUE)

Liles

Perfect - but I added "scale_color_discrete(guide = FALSE)" to remove the now unnecessary legends from the outside of the chart (saving some important screen real estate) – Anlage

Hello, can you expand it to this case: #48488213 ? – Crockett

A bit late to this, but I noticed the "a" in the State legend. Any way to have the original legend? With the "-" regarding the respective color (label) in the legend? – Goof

Answer 3

I provide another answer for weary ggplot folk.

This solution's principle can be applied quite generally.

Plot_df <- 
  temp.dat %>% mutate_if(is.factor, as.character) %>%  # Who has time for factors..
  mutate(Year = as.numeric(Year))

And now, we can subset our data

ggplot() + 
geom_line(data = Plot_df, aes(Year, Capex, color = State)) +
geom_text(data = Plot_df %>% filter(Year == last(Year)), aes(label = State, 
                                                           x = Year + 0.5, 
                                                           y = Capex, 
                                                           color = State)) + 
          guides(color = FALSE) + theme_bw() + 
          scale_x_continuous(breaks = scales::pretty_breaks(10))

The last pretty_breaks part is just to fix the axis below.

Jaenicke

Answer 4

There is a new package to address this very popular problem. {geomtextpath} gives some very flexible options for direct labelling, more than "only" labelling at the end...

Moreover, the labels will follow the curves! This might not be to everyone's taste, but I find this an extremely neat look.

library(geomtextpath)

## end of line
ggplot(temp.dat) +
  geom_textline(aes(
    x = Year, y = Capex, group = State, colour = State, label = State
  ),
  hjust = 1
  ) +
  theme(legend.position = "none")

## somewhere in the middle
ggplot(temp.dat) +
  geom_textline(aes(
    x = Year, y = Capex, group = State, colour = State, label = State
  ),
  hjust = .7
  ) +
  theme(legend.position = "none")

There are plenty of geoms and also one for prediction curves based on geom_smooth. (answering to user Mark Neal)

ggplot(temp.dat, aes(x = Year, y = Capex, group = State, colour = State)) +
  geom_line() +
  ## note this is using the current dev version. you currently have to specify method argument, otherwise the disambiguation of some function fails. 
  ## see also https://github.com/AllanCameron/geomtextpath/issues/79) +
  geom_textsmooth(aes(label = State), 
                  lty = 2, 
                  hjust = 1) +
  theme(legend.position = "none")
#> `geom_smooth()` using method = 'loess' and formula 'y ~ x'

^{Created on 2022-07-12 by the reprex package (v2.0.1)}

Chaddie

I haven't looked at geomtextpath in detail, but would love to see an example where it labels with a fitted line (e.g. end of fitted line) similar to my long hand answer (https://mcmap.net/q/103654/-plot-labels-at-ends-of-lines) – Glittery

@MarkNeal this is v easy with this package, see update – Chaddie

Answer 5

I'd like to add a solution for cases when you have longer label names. In all of the solutions provided, the labels are within the plot canvas, but if you have longer names, they'll get cut off. Here's how I solved that issue:

library(tidyverse)

# Make the "State" variable have longer levels
temp.dat <- temp.dat %>% 
    mutate(State = paste0(State, '-a-long-string'))

ggplot(temp.dat, aes(x = Year, y = Capex, color = State, group = State)) + 
    geom_line() +
    # Add labels at the end of the line
    geom_text(data = filter(temp.dat, Year == max(Year)),
              aes(label = State),
              hjust = 0, nudge_x = 0.1) +
    # Allow labels to bleed past the canvas boundaries
    coord_cartesian(clip = 'off') +
    # Remove legend & adjust margins to give more space for labels
    # Remember, the margins are t-r-b-l
    theme(legend.position = 'none',
          plot.margin = margin(0.1, 2.6, 0.1, 0.1, "cm")) 

Doyledoyley

I would appreciate if you help me if I want to put some mark or X or coordinates at only one point which is lower (in this case, how can I put some label at (2008, 5) in the WA-a-long-string? I will appreciate for your response! – Empiric

In my solution, I filter the data to select the exact x and y coordinates I want for my labels. Since I wanted them at the ends of the lines, I used data = filter(temp.dat, Year == max(Year)) inside the geom_text() call. In your case, you could change the filter to data = filter(temp.dat, Year == 2008, State = "WA"), which would give you only the "WA" label at the x position of 2008, and you could adjust the y position by adjusting the nudge_y parameter in geom_text() – Doyledoyley

I don't view this as an improvement as hard setting margins is not practical. Following my solution below: temp.dat <- temp.dat %>% mutate(State = paste0(State, '-a-long-string')) Plot_df <- temp.dat %>% mutate_if(is.factor, as.character) %>% mutate(Year = as.numeric(Year)) ggplot() + geom_line(data = Plot_df, aes(Year, Capex, color = State)) + geom_text(data = Plot_df %>% filter(Year == last(Year)), aes(label = State, x = Year + 3, y = Capex, color = State), hjust = 1) + guides(color = FALSE) + theme_bw() + scale_x_continuous(breaks = scales::pretty_breaks(10)) – Jaenicke

Not really sure what makes hard setting margins any less practical than hard setting scale limits. The top-ranked solution modifies the plot margins. The bigger difference I see between my solution and yours is that in my solution the x-axis stops at the last data point whereas in yours it continues out as far as needed such that the label name fits inside the plot boundary where there are no data points. – Doyledoyley

@Doyledoyley Two things. First, hard setting margins is not simpler than increasing the x margin (one input - 3 years, which is intuitive and simple. Margins are 4 inputs and unintuitive). To your last point, you want the figure's x-axis to expand - otherwise your names go outside your themes as in your solution (exactly what you don't want). In my solution - the name is still within your theme, yours go outside. That is not ideal of course. Top rated solution is dated (much more onerous than other solutions here) - and also has names go outside the theme selection. – Jaenicke

I guess at this point it's to each their own. If you're okay with your labels extending past the plot boundary in order to keep the x-axis limited to where there are data, then you have my solution. On the other hand, if you're okay with extending the x-axis beyond where there are data in order to keep the labels in the plot boundary, then you have your solution. – Doyledoyley

Answer 6

Not sure if it is the best way, but you could try the following (play a bit with xlim for correctly setting the limits):

library(dplyr)
lab <- tapply(temp.dat$Capex, temp.dat$State, last)
ggplot(temp.dat) + 
    geom_line(aes(x = Year, y = Capex, group = State, colour = State)) +
    scale_color_discrete(guide = FALSE) +
    geom_text(aes(label = names(lab), x = 12, colour = names(lab), y = c(lab), hjust = -.02))

Ultramontanism

This produces an error message: "Error: Aesthetics must be either length 1 or the same as the data (48): x, y, label, hjust" – Nolanolan

Answer 7

I came to this question looking to direct label a fitted line (e.g. loess()) at the last fitted point, not the last data point. I eventually worked out an approach to do this, largely based on tidyverse It should also work for linear regression with a few mods, so I leave it here for posterity.

library(tidyverse)

temp.dat$Year <- as.numeric(temp.dat$Year)
temp.dat$State <- as.character(temp.dat$State)

#example of loess for multiple models
#https://mcmap.net/q/103275/-loess-regression-on-each-group-with-dplyr-group_by

models <- temp.dat %>%
  tidyr::nest(-State) %>%
  dplyr::mutate(
    # Perform loess calculation on each CpG group
    m = purrr::map(data, loess,
                   formula = Capex ~ Year, span = .75),
    # Retrieve the fitted values from each model
    fitted = purrr::map(m, `[[`, "fitted")
  )

# Apply fitted y's as a new column
results <- models %>%
  dplyr::select(-m) %>%
  tidyr::unnest()

#find final x values for each group
my_last_points <- results %>% group_by(State) %>% summarise(Year = max(Year, na.rm=TRUE))

#Join dataframe of predictions to group labels
my_last_points$pred_y <- left_join(my_last_points, results)

# Plot with loess line for each group
ggplot(results, aes(x = Year, y = Capex, group = State, colour = State)) +
  geom_line(alpha = I(7/10), color="grey", show.legend=F) +
  #stat_smooth(size=2, span=0.3, se=F, show_guide=F)
  geom_point(size=1) +
  geom_smooth(se=FALSE)+
  geom_text(data = my_last_points, aes(x=Year+0.5, y=pred_y$fitted, label = State))

Glittery

Answer 8

You didn't emulate @Baptiste's solution 100%. You need to use annotation_custom and loop through all your Capex's:

library(ggplot2)
library(dplyr)
library(grid)

temp.dat <- structure(list(Year = c("2003", "2004", "2005", "2006", "2007", 
"2008", "2009", "2010", "2011", "2012", "2013", "2014", "2003", 
"2004", "2005", "2006", "2007", "2008", "2009", "2010", "2011", 
"2012", "2013", "2014", "2003", "2004", "2005", "2006", "2007", 
"2008", "2009", "2010", "2011", "2012", "2013", "2014", "2003", 
"2004", "2005", "2006", "2007", "2008", "2009", "2010", "2011", 
"2012", "2013", "2014"), State = structure(c(1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 
4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L), .Label = c("VIC", 
"NSW", "QLD", "WA"), class = "factor"), Capex = c(5.35641472365348, 
5.76523240652641, 5.24727577535625, 5.57988239709746, 5.14246402568366, 
4.96786288162828, 5.493190785287, 6.08500616799372, 6.5092228474591, 
7.03813541623157, 8.34736513875897, 9.04992300432169, 7.15830329914056, 
7.21247045701994, 7.81373928617117, 7.76610217197542, 7.9744994967006, 
7.93734452080786, 8.29289899132255, 7.85222269563982, 8.12683746325074, 
8.61903784301649, 9.7904327253813, 9.75021175267288, 8.2950673974226, 
6.6272705639724, 6.50170524635367, 6.15609626379471, 6.43799637295979, 
6.9869551384028, 8.36305663640294, 8.31382617231745, 8.65409824343971, 
9.70529678167458, 11.3102788081848, 11.8696420977237, 6.77937303542605, 
5.51242844820827, 5.35789621712839, 4.38699327451101, 4.4925792218211, 
4.29934654081527, 4.54639175257732, 4.70040615159951, 5.04056109514957, 
5.49921208937735, 5.96590909090909, 6.18700407463007)), class = "data.frame", row.names = c(NA, 
-48L), .Names = c("Year", "State", "Capex"))

temp.dat$Year <- factor(temp.dat$Year)

color <- c("#8DD3C7", "#FFFFB3", "#BEBADA", "#FB8072")

gg <- ggplot(temp.dat) 
gg <- gg + geom_line(aes(x=Year, y=Capex, group=State, colour=State))
gg <- gg + scale_color_manual(values=color)
gg <- gg + labs(x=NULL)
gg <- gg + theme_bw()
gg <- gg + theme(legend.position="none")

states <- temp.dat %>% filter(Year==2014)

for (i in 1:nrow(states))  {
  print(states$Capex[i])
  print(states$Year[i])
  gg <- gg + annotation_custom(
    grob=textGrob(label=states$State[i], 
                    hjust=0, gp=gpar(cex=0.75, col=color[i])),
    ymin=states$Capex[i],
    ymax=states$Capex[i],
    xmin=states$Year[i],
    xmax=states$Year[i])
}    

gt <- ggplot_gtable(ggplot_build(gg))
gt$layout$clip[gt$layout$name == "panel"] <- "off"
grid.newpage()
grid.draw(gt)

(You'll want to change the yellow if you keep the white background.)

Exorcist

Answer 9

And yet another option not already addressed by the answers and which I use quite often is to use the so-called secondary axis trick which means to use a secondary or duplicated axis to add direct labels to the end of the lines.

The advantage of this approach is that we do not need to fiddle around with extending the limits or the margins or ... to make room for the labels as the chart will be automatically be re-sized to make room for the axis text and hence plays nicely even for long labels.

To color labels I use ggh4x::guide_axis_color as in vanilla ggplot2 it is not officially supported to pass a vector of colors to axis.text.y.right and other approaches, e.g. using ggtext require slightly more work.

library(ggplot2)

# Dataframe of breaks and labels
sec_y <- subset(
  temp.dat,
  Year == max(Year),
  select = c(Capex, State)
) |>
  setNames(c("breaks", "labels"))

ggplot(
  temp.dat,
  aes(x = Year, y = Capex, color = State, group = State)
) +
  geom_line() +
  scale_y_continuous(
    sec.axis = dup_axis(
      breaks = sec_y$breaks,
      labels = sec_y$labels,
      # Add colors
      guide = ggh4x::guide_axis_color(
        color = scales::hue_pal()(nrow(sec_y))
      )
    )
  ) +
  guides(color = "none") +
  # Get rid of secondary axis ticks and title
  theme(
    axis.ticks.y.right = element_blank(),
    axis.title.y.right = element_blank()
  )

# Make the "State" variable have longer levels
temp.dat <- temp.dat |>
  transform(State = paste0(State, "-a-long-string"))

sec_y <- subset(
  temp.dat,
  Year == max(Year),
  select = c(Capex, State)
) |>
  setNames(c("breaks", "labels"))
sec_y <- sec_y[order(sec_y$labels), ]
ggplot(
  temp.dat,
  aes(x = Year, y = Capex, color = State, group = State)
) +
  geom_line() +
  scale_y_continuous(
    sec.axis = dup_axis(
      breaks = sec_y$breaks,
      labels = sec_y$labels,
      guide = ggh4x::guide_axis_color(
        color = scales::hue_pal()(nrow(sec_y))
      )
    )
  ) +
  guides(color = "none") +
  theme(
    axis.ticks.y.right = element_blank(),
    axis.title.y.right = element_blank()
  )

Eufemiaeugen