How to draw both positive mirror bar graph in r? - r

Question

I would like to draw a graph similar to the image here:
I tried to find similar mirror bar graphs on google, but I could not find similar graph to the image above.
Tricky parts of the graph are that 1) both +ve and -ve y axis have positive values, and 2) both +ve and -ve y axis have different y-axis labellings.
Thank you in advance for your help.

Answer 1

This is as close as I could get so far to that graph.
It's really tricky.
The Y axis has to be positive on the negative side
On the negative side numbers have to look 5 times smaller because of the number on the Y axis being 5 times smaller [from 1 to 5 instead of 1 to 25]
uncertainty bars need to drawn
X labels are doubled
What I couldn't do:
set up the Y axis names in a proper manner, [if anyone knows and can help..!]
understand what a and b are and with which logic to place them [you need to explain this one better]
library(dplyr)
library(ggplot2)
# your data
n <- 100
set.seed(42)
df <- tibble(var1 = factor(rep(c("Mamou", "Crowley"), each = 8 * n), levels = c("Mamou", "Crowley"), ordered = TRUE),
var2 = factor(rep(c("RWW-M1", "RWW-M2", "RWW-C1", "RWW-C2"), each = 4* n), levels = c("RWW-M1", "RWW-M2", "RWW-C1", "RWW-C2"), ordered = TRUE),
var3 = factor(rep(rep(c("Shoot dry weight (g)", "Root dry weight (g)"), each = 2*n), 4), levels = c("Shoot dry weight (g)", "Root dry weight (g)"), ordered = TRUE),
varc = rep(rep(c("white", "black"), each = n), 8),
value = abs(c(
rnorm(2*n, mean = 5 , sd = 0.2),
rnorm(2*n, mean = 3 , sd = 0.04),
rnorm(2*n, mean = 15 , sd = 0.2),
rnorm(2*n, mean = 4 , sd = 0.04),
rnorm(2*n, mean = 5 , sd = 0.2),
rnorm(2*n, mean = 2.5, sd = 0.04),
rnorm(2*n, mean = 5 , sd = 0.2),
rnorm(2*n, mean = 2.5, sd = 0.04))))
# edit your data this way [a little trick to set bars up and down the line and make them look like 5 times bigger]
df <- df %>% mutate(value = if_else(var3 == "Root dry weight (g)", -value*5, value))
# calculate statistics you want to plot
df <- df %>%
group_by(var1, var2, var3, varc) %>%
summarise(mean = mean(value), min = min(value), max = max(value)) %>%
ungroup()
df %>%
ggplot(aes(x = var2)) +
# plot dodged bars
geom_col(aes(y = mean, fill = varc),
position = position_dodge(width = 0.75),
colour = "black", width = 0.5) +
# plot dodged errorbars
geom_errorbar(aes(ymin = min, ymax = max, group = varc),
position = position_dodge(width = 0.75), width = 0.2, size = 1) +
# make line on zero more visible
geom_hline(aes(yintercept = 0)) +
# set up colour of the bars, don't show legend
scale_fill_manual(values = c("white", "gray75"), guide = FALSE) +
# set up labels of y axis
# dont change positive, make negative look positive and 5 times smaller
# set up breaks every 5 [ggplot will calc labels after breaks]
scale_y_continuous(labels = function(x) if_else(x<0, -x/5, x),
breaks = function(x) as.integer(seq(x[1]-x[1]%%5, x[2]-x[2]%%5, 5))) +
# put labels and x axis on top
scale_x_discrete(position = "top") +
# set up var1 labels on top
facet_grid( ~ var1, space = 'free', scales = 'free') +
# show proper axis names
labs(x = "", y = "Root dry weight (g) Shoot dry weight (g)") +
# set up theme
theme_classic() +
theme(axis.line.x = element_blank(),
axis.ticks.x = element_blank(),
panel.grid = element_blank(),
# this is to put names of facet grid on top
strip.placement = 'outside',
# this is to remove background from labels on facet grid
strip.background = element_blank(),
# this is to make facets close to each other
panel.spacing.x = unit(0,"line"))

Answer 2

Something like this perhaps?
library(ggplot2)
df <- data.frame(x = rep(letters[1:3], each = 4),
y = c(2, -2, 3, -3, 4, -4, 5, -5, 2, -2, 3, -3),
dodgegroup = factor(rep(rep(1:2, each = 2), 3)))
ggplot(df, aes(x, y, fill = dodgegroup)) +
geom_col(position = position_dodge(width = 0.75),
colour = "black", width = 0.5) +
geom_hline(aes(yintercept = 0)) +
scale_fill_manual(values = c("white", "gray75")) +
scale_y_continuous(breaks = 0:10 - 5,
labels = c(5:0, 5 * 1:5)) +
theme_classic()
Created on 2020-08-07 by the reprex package (v0.3.0)

Answer 3

Try this. While the answer by Edo looks most like what you have asked for, this method does not need you to transform your data. However, the scale on both sides of the axis are the same.
Call geom_col twice but with - before the values for Root, then we use labels=abs to make both sides of the y-axis positive numbers:
Edit - fixed the y-axis
library(ggplot2)
df <- data.frame(x = rep(c("RWW-M1", "RWW-M2", "RWW-C1", "RWW-C2"), each = 2),
Shoot = c(5, 6, 7, 8, 4, 5, 5, 7),
Root = c(1, 2, 3, 4, 2, 3, 1, 2),
Condition = rep(c("control", "test"), each = 1))
p <- ggplot(df, aes(x=x, fill=Condition)) +
geom_col(aes(y=Shoot), position = position_dodge(width = 0.75), width = 0.5, colour = "black")+
geom_col(aes(y=-Root), position = position_dodge(width = 0.75), width = 0.5, colour = "black")+
geom_hline(aes(yintercept = 0)) +
scale_fill_manual(values = c("white", "gray75")) +
ylab("Root weight (g) / Shoot weight (g)")+
xlab("")+
scale_y_continuous(breaks = 0:15 - 5, labels=abs) +
theme_bw()
p