How to add a line in boxplot?

Question

I would like to add lines between "mean" in my boxplot.

My code:

library(ggplot2)
library(ggthemes)

Gp=factor(c(rep("G1",80),rep("G2",80)))
Fc=factor(c(rep(c(rep("FC1",40),rep("FC2",40)),2)))
Z <-factor(c(rep(c(rep("50",20),rep("100",20)),4)))
Y <- c(0.19 , 0.22 , 0.23 , 0.17 , 0.36 , 0.33 , 0.30 , 0.39 , 0.35 , 0.27 , 0.20 , 0.22 , 0.24 , 0.16 , 0.36 , 0.30 , 0.31 , 0.39 , 0.33 , 0.25 , 0.23 , 0.13 , 0.16 , 0.18 ,  0.20 , 0.16 , 0.15 , 0.09 , 0.18 , 0.21 , 0.20 , 0.14 , 0.17 , 0.18 , 0.22 , 0.16 , 0.14 , 0.11 , 0.18 , 0.21 , 0.30 , 0.36 , 0.40 , 0.42 , 0.26 , 0.23 , 0.25 , 0.30 ,  0.27 , 0.15 , 0.29 , 0.36 , 0.38 , 0.42 , 0.28 , 0.23 , 0.26 , 0.29 , 0.24 , 0.17 , 0.24 , 0.14 , 0.17 , 0.16 , 0.15 , 0.21 , 0.19 , 0.15 , 0.16 , 0.13 , 0.25 , 0.12 ,  0.15 , 0.15 , 0.14 , 0.21 , 0.20 , 0.13 , 0.14 , 0.12 , 0.29 , 0.29 , 0.29 , 0.24 , 0.21 , 0.23 , 0.25 , 0.33 , 0.30 , 0.27 , 0.31 , 0.27 , 0.28 , 0.25 , 0.22 , 0.23 , 0.23 , 0.33 , 0.29 , 0.28 , 0.12 , 0.28 , 0.22 , 0.19 , 0.22 , 0.14 , 0.15 , 0.15 , 0.21 , 0.25 , 0.11 , 0.27 , 0.22 , 0.17 , 0.21 , 0.15 , 0.16 , 0.15 , 0.20 , 0.24 ,  0.24 , 0.25 , 0.36 , 0.24 , 0.34 , 0.22 , 0.27 , 0.26 , 0.23 , 0.28 , 0.24 , 0.23 , 0.36 , 0.23 , 0.35 , 0.21 , 0.25 , 0.26 , 0.23 , 0.28 , 0.24 , 0.23 , 0.09 , 0.16 , 0.16 , 0.14 , 0.18 , 0.18 , 0.18 , 0.12 , 0.22 , 0.23 , 0.09 , 0.17 , 0.15 , 0.13 , 0.17 , 0.19 , 0.17 , 0.11)
X <- factor(c(rep(c(rep("B1",10),rep("B2",10)),8)))
DATA=data.frame(Y,X,Z,Fc,Gp)
p <- qplot(X, Y, data=DATA, geom="boxplot", fill=Z, na.rm = TRUE, 
                    outlier.size = NA, outlier.colour = NA)  +
          facet_grid(Gp ~ Fc)+ theme_light()+scale_colour_gdocs()+
          theme(legend.position="bottom") + 
          stat_summary(fun.y=mean, geom="point", shape=23, position = position_dodge(width = .75))

I have:

And the expected plot I want:

I tried this

p + stat_summary(fun.y=mean, geom="line", aes(group = factor(Z)))

and this

p + stat_summary(fun.y=mean, geom="line", aes(group = factor(X)))

but none of the above worked. Instead, I received the following error message:

geom_path: Each group consists of only one observation. Do you need to adjust the group aesthetic? geom_path: Each group consists of only one observation. Do you need to adjust the group aesthetic? geom_path: Each group consists of only one observation. Do you need to adjust the group aesthetic? geom_path: Each group consists of only one observation. Do you need to adjust the group aesthetic?

Thanks for your help !

Scarce

Answer 1

You can try a tidyverse solution as well:

library(tidyverse)
DATA %>% 
   ggplot() + 
   geom_boxplot(aes(X, Y, fill=Z)) +
   stat_summary(aes(X, Y,fill=Z),fun.y = mean, geom = "point",
                position=position_nudge(x=c(-0.185,0.185))) +
   geom_segment(data=. %>%
                  group_by(X, Z, Gp , Fc) %>% 
                  summarise(M=mean(Y)) %>% 
                  ungroup() %>% 
                  mutate(Z=paste0("C",Z)) %>% 
                  spread(Z, M), aes(x = as.numeric(X)-0.185, y = C100, 
                    xend = as.numeric(X)+0.185, yend = C50)) +
   facet_grid(Gp ~ Fc)

The idea is the same as in the answer of d.b.. Create a data.frame for the geom_segment call. the advantage is the dplyr workflow. So everything is done in one run.

DATA %>% 
  group_by(X, Z, Gp , Fc) %>% 
  summarise(M=mean(Y)) %>% 
  ungroup() %>% 
  mutate(Z=paste0("C",Z)) %>% 
  spread(Z, M) 
# A tibble: 8 x 5
       X     Gp     Fc  C100   C50
* <fctr> <fctr> <fctr> <dbl> <dbl>
1     B1     G1    FC1 0.169 0.281
2     B1     G1    FC2 0.170 0.294
3     B1     G2    FC1 0.193 0.270
4     B1     G2    FC2 0.168 0.269
5     B2     G1    FC1 0.171 0.276
6     B2     G1    FC2 0.161 0.292
7     B2     G2    FC1 0.188 0.269
8     B2     G2    FC2 0.163 0.264

Or you can try a slighlty different approach compared to Julius' answer. Add breaks and labels to get the expected output and play around with some offset on a numeric X2 and the width parameter within the boxplot function to get the boxes plotted together.

DATA %>% 
  mutate(X2=as.numeric(interaction(Z, X))) %>% 
  mutate(X2=ifelse(Z==100, X2 + 0.2, X2 - 0.2)) %>% 
  ggplot(aes(X2, Y, fill=Z, group=X2)) + 
   geom_boxplot(width=0.6) +
   stat_summary(fun.y = mean, geom = "point") +
   stat_summary(aes(group = X),fun.y = mean, geom = "line") +
   facet_grid(Gp ~ Fc) +
   scale_x_continuous(breaks = c(1.5,3.5), labels = c("B1","B2"),
                        minor_breaks = NULL, limits=c(0.5,4.5))

Dispersion

Thank you @Jimbou for your help and your clear answer ! but I don't understand why you add X2+0.2 and X2-0.2 and breaks = c (1.5,3.5) ? – Scarce

The breaks = c (1.5,3.5) were set as you wanted a grouped boxplot of B1 and of B2. Thus this is the value inbetween. The offset of +/-0.2 together with the width=0.6 were chosen as then the boxes for F100 and F50 are closely plotted side by side. – Dispersion

@Sh.student Omit the scale_x_continuous part, then it should be clear. – Dispersion

with your code, can we eliminate or not display the outliers ? using for example : outlier.size = NA, outlier.colour = NA ... – Scarce

Yes of course. Add outlier.color = NA to the boxplot function. – Dispersion

Let us continue this discussion in chat. – Scarce

Answer 2

Here's an alternative:

DATA$U <- paste(X, Z) # Extra interaction
qplot(U, Y, data = DATA, geom = "boxplot", fill = Z, na.rm = TRUE, 
      outlier.size = NA, outlier.colour = NA) +
  facet_grid(Gp ~ Fc) + theme_light() + scale_colour_gdocs() +
  theme(legend.position = "bottom") + 
  stat_summary(fun.y = mean, geom = "point", shape = 23, position = position_dodge(width = .75)) +
  stat_summary(fun.y = mean, geom = "line", aes(group = X)) + # Lines
  scale_x_discrete(labels = rep(levels(X), each = 2)) + xlab("X") # Some fixes

Expediency

Good alternative ! but on the x-axis I want to have only B1 and B2 no B1 B1 and B2 B2, I mean, I want to grouped both boxplot of B1 and of B2, as I published in my question. Thank you for your help ! – Scarce

@Sh.student, I understand, but unfortunately that's the side effect of this approach. – Expediency

@Sh.student, https://mcmap.net/q/906337/-connect-ggplot-boxplots-using-lines-and-multiple-factor/1320535 is almost a duplicate and the second answer gives more details on my approach. – Expediency

Answer 3

This is not elegant but try this

tmp1 = aggregate(Y~., DATA[DATA$Z == 100,], mean)
tmp2 = aggregate(Y~., DATA[DATA$Z == 50,], mean)
tmp1$X2 = tmp2$X
tmp1$Y2 = tmp2$Y

graphics.off()
ggplot(DATA, aes(x = factor(X), y = Y, fill = Z)) +
    geom_boxplot(width = 0.5, outlier.shape = NA) +
    geom_segment(data = tmp1,
                 aes(x = as.numeric(factor(X)) - 0.125, y = Y,
                     xend = as.numeric(factor(X2)) + 0.125, yend = Y2)) +
    facet_grid(Gp ~ Fc)

Kirsten

Answer 4

Another approach, admittedly a bit convoluted, but hopefully it avoids some hardcoding.

The idea is to build a plot object including the stat_summary call. From this, grab relevant data (ggplot_build(p)$data[[2]]) to be used for the lines. The second data slot ([[2]]) corresponds to the second layer in the plot call, i.e. the x and y generated by stat_summary.

Grab x and y positions and indices of panel (PANEL) and x categories (group).

In the data from the plot object, the 'PANEL' and 'group' variables are not given explicitly by their names, but as numbers corresponding to the different combinations of facet variables, and variables which eventually will generate a numeric x position (here both 'the real' x and fill).

However, because categorical variables are ordered lexicographically in ggplot, we can match the numbers with their corresponding variables. The .GRP function in data.table is convenient here.

This data can then be used to draw a geom_line between the means.

# dodge value
pos <- position_dodge(width = 0.75)

# initial plot
p <- ggplot(data = DATA, aes(x = X, y = Y, fill = Z)) +
  geom_boxplot(outlier.size = NA, outlier.colour = NA, 
               position = pos) +
  stat_summary(fun.y = mean, geom = "point", shape = 23, position = pos) +
  facet_grid(Gp ~ Fc)

# grab relevant data
d <- ggplot_build(p)$data[[2]][ , c("PANEL", "group", "x", "y")]

library(data.table)
setDT(DATA)

# select unique combinations of facet and x variables
# here x includes the fill variable 'Z'
d2 <- unique(DATA[ , .(Gp, Fc, Z, X)])

# numeric index of facet combinations
d2[ , PANEL := .GRP, by = .(Gp, Fc)]

# numeric index of x combinations
d2[ , group := .GRP, by = .(Z, X)]

# add x and y positions by joining on PANEL and group
d2 <- d2[d, on = .(PANEL, group)]

# plot!
p + geom_line(data = d2, aes(x = x, y = y))

Bathsheeb

Answer 5

I have a way of doing this, surely similar to whats have been done, but using geom_line and position_dodge and data.table

library(data.table)
DATA=data.table(Y,X,Z,Fc,Gp)

 qplot(X, Y, data=DATA, geom="boxplot", fill=Z, na.rm = TRUE, 
           outlier.size = NA, outlier.colour = NA)  +
   geom_line(data = DATA[,list(Y = mean(Y)), by = .(X,Z,Fc,Gp)][X == "B1"],aes(X,Y,color = Z),group =1, position = position_dodge(width = .75),color = "black") +
   geom_line(data = DATA[,list(Y = mean(Y)), by = .(X,Z,Fc,Gp)][X == "B2"],aes(X,Y,color = Z),group =1, position = position_dodge(width = .75),color = "black") +
  facet_grid(Gp ~ Fc)+ theme_light()+
  theme(legend.position="bottom") +
  stat_summary(fun.y=mean, geom="point", shape=23, position = position_dodge(width = .75))

Cocoa