ggplot2: reorder bars from highest to lowest in each facet [duplicate] - r

Question

This question already has answers here:
ggplot bar plot with facet-dependent order of categories
(4 answers)
Closed 5 years ago.
In the df below, I want to reorder bars from highest to lowest in each facet
I tried
df <- df %>% tidyr::gather("var", "value", 2:4)
ggplot(df, aes (x = reorder(id, -value), y = value, fill = id))+
geom_bar(stat="identity")+facet_wrap(~var, ncol =3)
It gave me
It didn't order the bars from highest to lowest in each facet.
I figured out another way to get what I want. I had to plot each variable at a time, then combine all plots using grid.arrange()
#I got this function from #eipi10's answer
#http://stackoverflow.com/questions/38637261/perfectly-align-several-plots/38640937#38640937
#Function to extract legend
# https://github.com/hadley/ggplot2/wiki/Share-a-legend-between-two-ggplot2-graphs
g_legend<-function(a.gplot) {
tmp <- ggplot_gtable(ggplot_build(a.gplot))
leg <- which(sapply(tmp$grobs, function(x) x$name) == "guide-box")
legend <- tmp$grobs[[leg]]
return(legend)
}
p1 <- ggplot(df[df$var== "A", ], aes (x = reorder(id, -value), y = value, fill = id))+
geom_bar(stat="identity") + facet_wrap(~var, ncol =3)
fin_legend <- g_legend(p1)
p1 <- p1 + guides(fill= F)
p2 <- ggplot(df[df$var== "B", ], aes (x = reorder(id, -value), y = value, fill = id))+
geom_bar(stat="identity") + facet_wrap(~var, ncol =3)+guides(fill=FALSE)
p3 <- ggplot(df[df$var== "C", ], aes (x = reorder(id, -value), y = value, fill = id))+
geom_bar(stat="identity") + facet_wrap(~var, ncol =3)+guides(fill=FALSE)
grid.arrange(p1, p2, p3, fin_legend, ncol =4, widths = c(1.5, 1.5, 1.5, 0.5))
The result is what I want
I wonder if there is a straightforward way that can help me order the bars from highest to lowest in all facets without having to plot each variable separtely and then combine them. Any suggestions will be much appreciated.
DATA
df <- read.table(text = c("
id A B C
site1 10 15 20
site2 20 10 30
site3 30 20 25
site4 40 35 40
site5 50 30 35"), header = T)

Answer 1

The approach below uses a specially prepared variable for the x-axis with facet_wrap() but uses the labels parameter to scale_x_discrete() to display the correct x-axis labels:
Prepare data
I'm more fluent in data.table, so this is used here. Feel free to use what ever package you prefer for data manipulation.
Edit: Removed second dummy variable, only ord is required
library(data.table)
# reshape from wide to long
molten <- melt(setDT(df), id.vars = "id")
# create dummy var which reflects order when sorted alphabetically
molten[, ord := sprintf("%02i", frank(molten, variable, -value, ties.method = "first"))]
molten
# id variable value ord
# 1: site1 A 10 05
# 2: site2 A 20 04
# 3: site3 A 30 03
# 4: site4 A 40 02
# 5: site5 A 50 01
# 6: site1 B 15 09
# 7: site2 B 10 10
# 8: site3 B 20 08
# 9: site4 B 35 06
#10: site5 B 30 07
#11: site1 C 20 15
#12: site2 C 30 13
#13: site3 C 25 14
#14: site4 C 40 11
#15: site5 C 35 12
Create plot
library(ggplot2)
# `ord` is plotted on x-axis instead of `id`
ggplot(molten, aes(x = ord, y = value, fill = id)) +
# geom_col() is replacement for geom_bar(stat = "identity")
geom_col() +
# independent x-axis scale in each facet,
# drop absent factor levels (not the case here)
facet_wrap(~ variable, scales = "free_x", drop = TRUE) +
# use named character vector to replace x-axis labels
scale_x_discrete(labels = molten[, setNames(as.character(id), ord)]) +
# replace x-axis title
xlab("id")
Data
df <- read.table(text = "
id A B C
site1 10 15 20
site2 20 10 30
site3 30 20 25
site4 40 35 40
site5 50 30 35", header = T)

Answer 2

If you're willing to lose the X axis labels, you can do this by using the actual y values as the x aesthetic, then dropping unused factor levels in each facet:
ggplot(df, aes (x = factor(-value), y = value, fill = id))+
geom_bar(stat="identity", na.rm = TRUE)+
facet_wrap(~var, ncol =3, scales = "free_x", drop = TRUE) +
theme(
axis.text.x = element_blank(),
axis.ticks.x = element_blank()
)
Result:
The loss of x-axis labels is probably not too bad here as you still have the colours to go on (and the x-axis is confusing anyway since it's not consistent across facets).