Put stars on ggplot barplots and boxplots - to indicate the level of significance (p-value)

It's common to put stars on barplots or boxplots to show the level of significance (p-value) of one or between two groups, below are several examples:

The number of stars are defined by p-value, for example one can put 3 stars for p-value < 0.001, two stars for p-value < 0.01, and so on (although this changes from one article to the other).

And my questions: How to generate similar charts? The methods that automatically put stars based on significance level are more than welcome.

Please find my attempt below.

First, I created some dummy data and a barplot which can be modified as we wish.

windows(4,4)

dat <- data.frame(Group = c("S1", "S1", "S2", "S2"),
                  Sub   = c("A", "B", "A", "B"),
                  Value = c(3,5,7,8))  

## Define base plot
p <-
ggplot(dat, aes(Group, Value)) +
    theme_bw() + theme(panel.grid = element_blank()) +
    coord_cartesian(ylim = c(0, 15)) +
    scale_fill_manual(values = c("grey80", "grey20")) +
    geom_bar(aes(fill = Sub), stat="identity", position="dodge", width=.5)

Adding asterisks above a column is easy, as baptiste already mentioned. Just create a data.frame with the coordinates.

label.df <- data.frame(Group = c("S1", "S2"),
                       Value = c(6, 9))

p + geom_text(data = label.df, label = "***")

To add the arcs that indicate a subgroup comparison, I computed parametric coordinates of a half circle and added them connected with geom_line. Asterisks need new coordinates, too.

label.df <- data.frame(Group = c(1,1,1, 2,2,2),
                       Value = c(6.5,6.8,7.1, 9.5,9.8,10.1))

# Define arc coordinates
r <- 0.15
t <- seq(0, 180, by = 1) * pi / 180
x <- r * cos(t)
y <- r*5 * sin(t)

arc.df <- data.frame(Group = x, Value = y)

p2 <-
p + geom_text(data = label.df, label = "*") +
    geom_line(data = arc.df, aes(Group+1, Value+5.5), lty = 2) +
    geom_line(data = arc.df, aes(Group+2, Value+8.5), lty = 2)

Lastly, to indicate comparison between groups, I built a larger circle and flattened it at the top.

r <- .5
x <- r * cos(t)
y <- r*4 * sin(t)
y[20:162] <- y[20] # Flattens the arc

arc.df <- data.frame(Group = x, Value = y)

p2 + geom_line(data = arc.df, aes(Group+1.5, Value+11), lty = 2) +
     geom_text(x = 1.5, y = 12, label = "***")