A quick and dirty (some hard-coding) possibility is to use geom_rect
, where the positions are given by the numerical values of the levels of x and y variables to be bound with a box, plus/minus an offset.
ggplot(mymelt, aes(x = industry, y = variable, fill = value, label = value)) +
geom_tile() +
geom_text() +
geom_rect(aes(xmin = 1 - 0.5, xmax = 2 + 0.5, ymin = 2 - 0.5, ymax = 3 + 0.5),
fill = "transparent", color = "red", size = 1.5)
A less hard-coded version:
# convert x and y variables to factors
ind <- as.factor(mymelt$industry)
vars <- as.factor(mymelt$variable)
# numeric version of the levels to be bound by a box
xmin <- unique(as.numeric(ind[ind == "all industries"]))
xmax <- unique(as.numeric(ind[ind == "cars"]))
ymin <- unique(as.numeric(vars[vars == "americas"]))
ymax <- unique(as.numeric(vars[vars == "europe"]))
# set offset
offset <- 0.5
ggplot(mymelt, aes(x = industry, y = variable, fill = value, label = value)) +
geom_tile() +
geom_text() +
geom_rect(aes(xmin = xmin - offset,
xmax = xmax + offset,
ymin = ymin - offset,
ymax = ymax + offset),
fill = "transparent", color = "red", size = 1.5)
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…