R ggplot2: How can I name the y axis depending on the value of a variable with facet_wrap?

Question

I'll give you an idea of the data and I think then it should be easier to understand what I'm trying to achieve.

Repex:

ID <- c(1, 1, 2, 3, 3, 3)
cat <- c("Others", "Others", "Population", "Percentage", "Percentage", "Percentage")
logT <- c(2.7, 2.9, 1.5, 4.3, 3.7, 3.3)
m <- c(1.7, 1.9, 1.1, 4.8, 3.2, 3.5)
aggr <- c("median", "median", "geometric mean", "mean", "mean", "mean")
over.under <- c("overestimation", "overestimation", "underestimation", "underestimation", "underestimation", "underestimation")
data <- cbind(ID, cat, logT, m, aggr, over.under)
data <- data.frame(data)
data$ID <- as.numeric(data$ID)
data$logT<- as.numeric(data$logT)
data$m<- as.numeric(data$m)

Code:

Fig <- data %>% ggplot(aes(x = logT, y = m, color = over.under)) + 
  facet_wrap(~ ID) +
  geom_point() +
  scale_x_continuous(name = "log (True value)", limits=c(1, 7)) +
  scale_y_continuous(name = NULL, limits=c(1, 7)) +
  geom_abline(intercept = 0, slope = 1, linetype = "dashed") +
  theme_bw() +
  theme(legend.position='none')
Fig

I want to label the y axis of each graph with the value of aggr. So for ID 1 it should say median, for ID 2 geometric mean and ID 3 mean.

I tried multiple things:

mtext(data1$aggr, side = 2, cex=1) #or
ylab(data1$aggr) #or
strip.position = "left"

But it doesn't work.

I'm also trying to add the cat in the top left corner of the graph. So for ID 1 "Others", ID 2 "Population" and ID 3 "Percentage". I tried to work with legend() but I have not been able to solve the problem yet either.

Answer 1

mtext is meant for plot(). ggplot is another plotting system so it will not work. Not many choices unfortunately, one way is to remove the xlab, and use the strip as a y-axis:

LAB =tapply(as.character(data$aggr),data$ID,unique)

Fig <- data %>% ggplot(aes(x = logT, y = m, color = over.under)) + 
  geom_point() +
  scale_x_continuous(name = "log (True value)", limits=c(1, 7)) +
  scale_y_continuous(name = NULL, limits=c(1, 7)) +
  geom_abline(intercept = 0, slope = 1, linetype = "dashed") +
  theme_bw() +
  theme(legend.position='none') +
  facet_wrap(~ID, scales = "free_y",strip.position = "left", 
  labeller = as_labeller(LAB ))  +
  ylab(NULL) +
  theme(strip.background = element_blank(),strip.placement = "outside")

The other way is to combine plots:

library(gridExtra) 

plts = by(data,data$ID,function(i){
ggplot(i,aes(x=logT,y=m,color=over.under)) + 
geom_point() + 
scale_x_continuous(name = "log (True value)", limits=c(1, 7)) +
scale_y_continuous(name = unique(i$agg), limits=c(1, 7)) +
geom_abline(intercept = 0, slope = 1, linetype = "dashed") +
theme_bw() + 
scale_color_manual(values=c("overestimation"="turquoise","underestimation"="orange"))+
theme(legend.position='none') 
})

grid.arrange(grobs=plts,ncol=3)

Answer 2

If we care about the ID facet labels, this gets a lot more complicated and is inspired by this answer.

First, we need to make two copies of the plot, one with the renamed strips and one with the original.

Then we add the facet strips manually to the other.

library(gtable)
library(grid)
plot1 <- data %>% ggplot(aes(x = logT, y = m, color = over.under)) + 
  facet_wrap(~ ID, scales = "free_y",strip.position = "left",  labeller = as_labeller(c(`1`="median",`2`="geometric mean",`3`="mean"))) +
  geom_point() +
  scale_x_continuous(name = "log (True value)", limits=c(1, 7)) +
  scale_y_continuous(name = NULL, limits=c(1, 7)) +
  geom_abline(intercept = 0, slope = 1, linetype = "dashed") +
  theme_bw() +  
  theme(legend.position='none',strip.background = element_blank(),strip.placement = "outside")

plot2 <- data %>% ggplot(aes(x = logT, y = m, color = over.under)) + 
  facet_grid(~ ID) +
  geom_point() +
  scale_x_continuous(name = "log (True value)", limits=c(1, 7)) +
  scale_y_continuous(name = NULL, limits=c(1, 7)) +
  geom_abline(intercept = 0, slope = 1, linetype = "dashed") +
  theme_bw() +  
  theme(legend.position='none')

gt1 = ggplot_gtable(ggplot_build(plot1))
gt2 = ggplot_gtable(ggplot_build(plot2))
strip1 <- gtable_filter(gt2, 'strip-t-1')
strip2 <- gtable_filter(gt2, 'strip-t-2')
strip3 <- gtable_filter(gt2, 'strip-t-3')
gt1 = gtable_add_rows(gt1, heights=strip1$heights[1], pos = 0)
panel_id <- gt1$layout[grep('panel-.+1$', gt1$layout$name),]
gt1 = gtable_add_grob(gt1, strip1, t = 1, l = panel_id$l[1])
gt1 = gtable_add_grob(gt1, strip2, t = 1, l = panel_id$l[2])
gt1 = gtable_add_grob(gt1, strip3, t = 1, l = panel_id$l[3])
gt1 = gtable_add_grob(gt1, zeroGrob(), t = 1, l = 1)
gt1 = gtable_add_rows(gt1, heights=gt2$heights[1], pos = 0)
grid.newpage()
grid.draw(gt1)