highlight points in ggplot2 stat_qq output - r

Question

I am trying to highlight selected points based on their order statistics in a ggplot stat_qq output:
ydata <- data.frame(sample = c(rep("Sample 1", 100),
rep("Sample 2", 100),
rep("Sample 3", 100),
rep("Sample 4", 100)),
x=rnorm(400))
ydata <- ydata[order(ydata$sample, ydata$x),]
ydata$order <- 1:100
ggplot(ydata, aes(sample=x)) + stat_qq() + facet_wrap(~sample) + scale_x_continuous(breaks = -2:2, labels = function(x) paste0(x, " \n [",100 * signif(pnorm(-2:2, lower.tail=FALSE),2), "%]")) + theme_bw(base_size = 14, base_family = "sans") + labs(title = "Four Samples of 100 Observations From Normal Distribution",
caption = "4 Samples of n = 100 from Normal Distribution \nNumbers indicate order of value",
y = "Sample Value",
x = "Standard Deviation\n[%exceeding]") +
geom_text(data = ydata[ydata$order %in% c(2,16,50,84,98),], aes(x=qnorm(pnorm(x)), y=x, label = order), nudge_y = 1)
Which produced this:
Obviously my text notation is not highlighting the right points (the 2, 16, 50 84, 98th points). I wish I could also highlight the actual points in red. Would appreciate any suggestions.

Answer 1

You could calculate the qq values outside of ggplot and create a separate column to group the qq values into highlighted and not highlighted. Then you could plot them using geom_point with the grouping variable as a colour aesthetic. For example:
library(tidyverse)
# Generate data reproducibly
set.seed(2)
ydata <- data.frame(sample = c(rep("Sample 1", 100),
rep("Sample 2", 100),
rep("Sample 3", 100),
rep("Sample 4", 100)),
x=rnorm(400))
ydata <- ydata[order(ydata$sample, ydata$x),]
ydata$order <- 1:100
# Quantile indices to highlight
pts = c(2,16,50,84,98)
# Add qq values and grouping column to data frame and pipe into ggplot
# Use split and map to calculate the qq values separately for each Sample
split(ydata, ydata$sample) %>%
map_df(~ .x %>% mutate(xq = qqnorm(x, plot.it=FALSE)$x,
group = ifelse(order %in% pts, "A", "B"))) %>%
ggplot(aes(xq, x, colour=group)) +
geom_point(size=1) +
geom_text(aes(label=ifelse(group=="A", order, "")),
nudge_y=1, size=3) +
facet_wrap(~ sample) +
theme_bw(base_size = 14, base_family = "sans") +
scale_colour_manual(values=c("red", "black")) +
guides(colour=FALSE)
As an alternative, a quick hack would be to use ggplot_build to highlight specific points in your original plot (note though that something is not quite right with how you placed the labels relative to the highlighted points):
pts = rep(c(2,16,50,84,98), 4) + rep(seq(0,300,100), each=5)
# Assuming you've assigned your plot to the object p
pb = ggplot_build(p)
# Change point colors
pb$data[[1]][pts, "colour"] = "red"
# Change label colors
pb$data[[2]][["colour"]] = "red"
# Regenerate plot object
p = ggplot_gtable(pb)
plot(p)

Answer 2

You can apply stat="qq" to your geom_point and then use the colors assigned to new variable
ydata <- data.frame(sample = c(rep("Sample 1", 100),
rep("Sample 2", 100),
rep("Sample 3", 100),
rep("Sample 4", 100)),
x=rnorm(400))
ydata <- ydata[order(ydata$sample, ydata$x),]
ydata$order <- 1:100
ydata$highlight = ifelse(ydata$order %in% c(2,16,50,84,98), "#FF0000", "#000000")
ydata$order_txt = ifelse(ydata$order %in% c(2,16,50,84,98), ydata$order, "")
ggplot(ydata, aes(sample=x)) +
geom_point(color=ydata$highlight, stat="qq") +
geom_text(label=ydata$order_txt, stat="qq", nudge_y=1) +
facet_wrap(~sample) +
scale_x_continuous(breaks = -2:2, labels = function(x) paste0(x, " \n [",100 * signif(pnorm(-2:2, lower.tail=FALSE),2), "%]")) +
theme_bw(base_size = 14, base_family = "sans") +
labs(
title = "Four Samples of 100 Observations From Normal Distribution",
caption = "4 Samples of n = 100 from Normal Distribution \nNumbers indicate order of value",
y = "Sample Value",x = "Standard Deviation\n[%exceeding]")