Highlight (shade) plot background in specific time range

Question

On a generic plot, with time on the x-axis, I would like to highlight a period of some specific years.

How can I bestly do this? My idea is for example a light yello

Answer 1

Using alpha transparency:

x <- seq(as.POSIXct("1949-01-01", tz="GMT"), length=36, by="months")
y <- rnorm(length(x))

plot(x, y, type="l", xaxt="n")
rect(xleft=as.POSIXct("1950-01-01", tz="GMT"),
     xright=as.POSIXct("1950-12-01", tz="GMT"),
     ybottom=-4, ytop=4, col="#123456A0") # use alpha value in col
idx <- seq(1, length(x), by=6)
axis(side=1, at=x[idx], labels=format(x[idx], "%Y-%m"))

or plot highlighted region behind lines:

plot(x, y, type="n", xaxt="n")
rect(xleft=as.POSIXct("1950-01-01", tz="GMT"),
     xright=as.POSIXct("1950-12-01", tz="GMT"),
     ybottom=-4, ytop=4, col="lightblue")
lines(x, y)
idx <- seq(1, length(x), by=6)
axis(side=1, at=x[idx], labels=format(x[idx], "%Y-%m"))
box()

Answer 2

You can use the chartSeries() function in quantmod with an xts timeSeries and the addTA() function to add the background highlighting:

addTA(xts(rep(TRUE,length(times)), times), on=-1, col="#333333", border=NA)

Answer 3

Here is a solution that uses zoo simply because that makes the subsetting easy. You could do the same with standard indexing as well:

## create a long monthly sequence and a sub-sequence
months <- seq( as.Date("1950-01-01"), as.Date("2009-12-12"), by="month")
subset <- seq( as.Date("1970-01-01"), as.Date("1979-12-31"), by="month")

## generate some random values
set.seed(42)
values <- cumsum(rnorm(length(months)))

## plot as a zoo object, overlay a gray background and overplot a line in red
library(zoo)
Z <- zoo(values, months)
plot(Z)
rect(xleft=head(subset,1), xright=tail(subset,1),
     ybottom=par("usr")[3], ytop=par("usr")[4],
     density=NA, col="lightgray")
lines(Z[subset], col='red')
box()

_{(source: eddelbuettel.com)}

By using par("usr") we avoid the need for explicit values for upper and lower region marks. And the zoo indexing makes finding the start- and endpoints easy. This would work the same way for data in different time resolutions.