Draw lines between two different ("grid.arranged") plots

Question

So i have managed to create two ggplot-graphs and plot them perfectly aligned with grid.arrange. Now i would like to connect the min, 1Q, 3Q and max of the boxplot with the distribution in the left graph. I have plotted horizontal (red striped) lines with geom_hline in both graphs, but i would like to have these lines connected, so also go through the white space between the graphs. Any suggestions? Here's the graph:

Oh yeah, complicating factor: the right graphs has it's coordinates flipped!

Here's a reproducible example:

library(ggplot2)
library(scales)
library(gridExtra)

# create some data
df_ahn <- data.frame(yh=rnorm(1000,0.23,0.05))
df_peil <- data.frame(hoogte = rnorm(1000,0,1))

# vector met de hoogtes
Summary_df <- summary(df_ahn$yh)

p_peil <- ggplot(df_peil,aes(x=hoogte))+
  geom_histogram(aes(y=cumsum((..count..)/sum(..count..))), binwidth = 0.01,fill="gray")+
  stat_bin(aes(y=cumsum((..count..)/sum(..count..))),binwidth = 0.01,geom="line",color="black") +
  geom_vline(aes(xintercept = as.vector(Summary_df[1])), lty = 2,color =2)+
  geom_vline(aes(xintercept = as.vector(Summary_df[2])), lty = 2,color =2)+
  geom_vline(aes(xintercept = as.vector(Summary_df[5])), lty = 2,color =2)+
  geom_vline(aes(xintercept = as.vector(Summary_df[6])), lty = 2,color =2)+
  coord_flip() +
  ggtitle("Onderschrijdingsfrequentie\n waterstand in kreek") +
  xlab("Hoogte in meter NAP") +
  ylab("Onderschrijdingsfrequentie in % (10% = 36,5 dagen/jaar)") +
  scale_y_continuous(limits = c(0, 1),labels = percent, breaks=seq(0,1,by=0.1)) +
  scale_x_continuous(limits = c(-0.5, 0.6), breaks=seq(-0.5,0.6,by=0.1)) +
  theme(panel.background = element_rect(fill = "transparent",colour = "black"),
        panel.grid.major = element_line(colour = "darkgray"),
        panel.grid.minor = element_line(colour = "gray"),
        strip.background = element_rect(fill="gray"),
        strip.text = element_text(size=14, color="black"),
        axis.ticks.y = element_line(colour = "black"),
        axis.ticks.x = element_line(colour = "black"),
        axis.text.x = element_text(size=14, color="black"),
        axis.text.y = element_text(size=14, color="black"),
        axis.title = element_text(size=14, color="black")
  )

p_ahn <-    ggplot(df_ahn, aes(x=1, y=yh)) +
  geom_boxplot(outlier.size=3, outlier.shape=1) +
  geom_hline(aes(yintercept = as.vector(Summary_df[1])), lty = 2,color =2)+
  geom_hline(aes(yintercept = as.vector(Summary_df[2])), lty = 2,color =2)+
  geom_hline(aes(yintercept = as.vector(Summary_df[5])), lty = 2,color =2)+
  geom_hline(aes(yintercept = as.vector(Summary_df[6])), lty = 2,color =2)+
  scale_y_continuous(limits = c(-0.5,0.6), breaks=seq(-0.5,0.6,by=0.1)) +
  ggtitle("Hoogte groeiplaatsen\nKruipend moerascherm") +
  ylab("") +
  xlab("") +
  theme(panel.background = element_rect(fill = "transparent",colour = "black"),
        panel.grid.major = element_blank(),
        panel.grid.minor = element_blank(),
        strip.background = element_rect(fill="gray"),
        strip.text = element_text(size=14, color="black"),
        axis.ticks.y = element_line(colour = "black"),
        axis.ticks.x = element_line(colour = "white"),
        axis.text.x = element_text(size=14, color="white"),
        axis.text.y = element_text(size=14, color="black"),
        axis.title = element_text(size=14, color="black", face="bold")

  ) 
grid.arrange(p_peil,p_ahn, layout_matrix = matrix(c(1,1,1,2), nrow=1, byrow=TRUE), ncol = 4)

[This Q&A](http://stackoverflow.com/questions/17492230/how-to-place-grobs-with-annotation-custom-at-precise-areas-of-the-plot-region) may be relevant. — Henrik, Aug 22 '15 at 16:41
or [this one](http://stackoverflow.com/questions/31690007/ggplot-drawing-line-between-points-across-facets/31691313#31691313). It seems to me that facetting would be the best option though. — baptiste, Aug 22 '15 at 21:23

baptiste · Accepted Answer · 2015-08-23T19:45:49.073

4

You could extract the line from one of the plots, and add it to the whole region of the combined gtable,

library(ggplot2)
library(gtable)
library(grid)

set.seed(123)
y <- rnorm(10)
p1 <- qplot(1:10, y) +
  geom_hline(yintercept=0, lty=3)

p2 <- qplot(1:10, 10*y) +
  geom_hline(yintercept=0)

#library(gridExtra)
#grid.arrange(p1,p2,widths=c(3,1)) # no line

g1 <- ggplotGrob(p1)
g2 <- ggplotGrob(p2)
lines <- g1$grobs[[4]][["children"]][[3]]
g1$grobs[[4]][["children"]][[3]] <- NULL # remove line
g2$grobs[[4]][["children"]][[3]] <- NULL # remove line
g <- cbind(g1,g2,size="first")
g$heights <- unit.pmax(g1$heights, g2$heights)
g$widths[[9]] <- unit(1/3, "null")
g <- gtable_add_grob(g, lines, l=4, t=3, r=9, z=Inf)
grid.newpage()
grid.draw(g)

edited Aug 23 '15 at 19:45

answered Aug 22 '15 at 21:47

baptiste

75,767
19
198
294

Thanks @baptiste for the response and the `widths`-TIP. However, your code gives the error **Error in mmm < each : comparison of these types is not implemented** after the `cbind`-line. I am not familiair with gtable, so can't figure out why. Any idea? BTW: I did provide sample data (and forgot to remove the irrelevant `setwd`-line). – RHA Aug 23 '15 at 12:38
sorry, this has always bugged me so I tweaked the cbind.gtable function in gridExtra, but it's only exported in the dev version. I've amended the code to work with the official gtable version. – baptiste Aug 23 '15 at 19:45
The line `g <- gtable_add_grob(g, lines, l=4, t=3, r=9, z=Inf)` throws an error: `Error: is.list(grobs) is not TRUE`, my package verisons are: `gtable_0.2.0`, `ggplot2_2.2.1`, `R 3.4.3` – snaut May 15 '18 at 10:28

Draw lines between two different ("grid.arranged") plots

1 Answers1

Linked