Program labels on line graph to avoid line ggplot2

Question

I'm curious if there is a way to overlay labels on a line graph on ggplot2 to avoid going through the line. I've used vjust, which works in most instances, but when there is a large increase or decrease between two dates the line goes through the label making it hard to read. I will put the plot I'm currently using below and the code. In this instance, I'm wanting to move the 920; 1,467; and 1,480 off the line. I'm exporting the plot to powerpoint via the Reporters package, so I can move it manually, I was just wondering if there was a way to avoid that.

Plot:

Code:

library("ggplot2")
library("scales")

line_data <- c(276, 475, 753, 840, 931, 962, 801, 920, 1467, 1840, 1737, 1638, 1789, 1733, 1480, 1464, 1538)
year_data <- c(2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016,
               2017)

line_data_total <- as.data.frame(cbind(line_data, year_data))

limit_func <- function(x) {
  if (nchar(max(x)) == 2){
    round(max(x +5), digits = -1)
  } else if (nchar(max(x)) == 3){
    round(max(x +50), digits = -2)
  } else if (nchar(max(x)) == 4){
    round(max(x +500), digits = -3)
  }
}

ggplot(data = line_data_total, aes(x = year_data, y = line_data, group = 1)) +
  geom_line(color = "red", size = 1.2)+
  geom_point(color = "red", fill = "red", shape = 23, size = 1.75) +
  geom_text(aes(label = paste0(format(round(as.numeric(line_data), 1), nsmall = 0, big.mark = ","))),
            size = 3, fontface = "bold", vjust = -2) +
  labs(x = '', y = '') +
  expand_limits(y = c(0, limit_func(line_data_total$line_data))) +
  scale_x_continuous(breaks = seq(min(line_data_total$year_data), max(line_data_total$year_data), 1)) +
  scale_y_continuous(labels = comma) +
  theme(panel.grid.major.x = element_blank() ,
        panel.grid.major.y = element_line( size=.1, color="light gray"),
        panel.background = element_rect(fill = "transparent"),
        plot.background = element_rect(fill = "transparent"),
        axis.text.x = element_text(face = "bold", size = 10),
        axis.text.y = element_text(face = "bold", size = 10),
        axis.ticks.y = element_blank())

Answer 1

How about using the ggplot2 extension ggrepel:

require(ggrepel)

Replace your geom_text() line with:

geom_text_repel(aes(label = paste0(format(round(as.numeric(line_data), 1), nsmall = 0, big.mark = ","))),
            size = 3, fontface = "bold", nudge_y=150)

nudge_y is what is pushing the labels off the line, you could use a combination of nudge_x and nudge_y for more control. And see the package vignette: https://github.com/slowkow/ggrepel/blob/master/vignettes/ggrepel.md

Answer 2

Without fiddling with the settings for geom_text_repel, I wanted to try a solution that was generalized for time series like this. I wrote a function that looks at the two x-neighboring points and calculates the slope between them and whether the middle point is lower, higher, or between its x-neighbors in the y direction.

• If a point is y-between its x-neighbors, the label will be shifted off diagonally, away from the slope of the line connecting them.
• If the point is y-below its x-neighbors, the label is centered, but shifted down.
• If the point is y-above its x-neighbors, the label is centered, but shifted up.

---

library(dplyr)

adjust_away_from_line <- function(df, x, y, vextend = 0.5) {
  if(!is.data.frame(df)) {return(df)}

  x <- enquo(x)
  y <- enquo(y)

  if(!(quo_name(x) %in% names(df))) {
    warning(paste0("Column '", quo_name(x), "' not found in data."))
    return(df)
  }

  if(!(quo_name(y) %in% names(df))) {
    warning(paste0("Column '", quo_name(y), "' not found in data."))
    return(df)
  }


  df %>% arrange(!!x) %>% 
    mutate(nb.slope = case_when(
      is.na(lead(!!y)) ~ (    (!!y) - lag(!!y))/(    (!!x) - lag(!!x)),
      is.na(lag(!!y))  ~ (lead(!!y) -    (!!y))/(lead(!!x) -    (!!x)),
      TRUE             ~ (lead(!!y) - lag(!!y))/(lead(!!x) - lag(!!x))  
    ),
    nb.pos = case_when(
      is.na(lead(!!y))                              ~ -sign(nb.slope),
      is.na(lag(!!y))                               ~ -sign(nb.slope),
       (lead(!!y) >= (!!y)) &  (lag(!!y) >= (!!y))  ~  1.1,
      !(lead(!!y) >= (!!y)) & !(lag(!!y) >= (!!y))  ~ -1.1,
      TRUE                                          ~ -1
    ),
    hjust = case_when(
      nb.pos   >  1 ~ 0.5,
      nb.pos   < -1 ~ 0.5,
      nb.slope >  0 ~ 1,
      nb.slope <  0 ~ 0,
      TRUE          ~ 0.5
    ),
    vjust = scales::rescale(round(nb.pos), to = c(0-vextend, 1+vextend))) %>% 
    select(-nb.slope, -nb.pos)


}

Since it takes a dataframe as its first argument, you can use this function in a pipe, giving it the bare names of your x variable and y variable in order:

data.frame(line_data = c(276, 475, 753, 840, 931, 962, 801, 920, 1467, 
                         1840, 1737, 1638, 1789, 1733, 1480, 1464, 1538),
           year_data = c(2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 
                         2011, 2012, 2013, 2014, 2015, 2016, 2017)
           ) %>%
  adjust_away_from_line(year_data, line_data) %>%
  ggplot(aes(year_data, line_data)) +
  geom_line() + 
  geom_point() +
  geom_text(aes(label = line_data, hjust = hjust, vjust = vjust))

If you want to move the labels further up and down away from the lines, you can do that with the adjust_away_from_line(..., vextend = ##) argument. The default is 0.5, but you might want 0.75 or 1 in different applications.