Yihui Xie

moving.block()

Cycle through an R object and plot each subset of elements

Yihui Xie & Lijia Yu / 2017-04-04


For a long numeric vector or matrix (or data frame), we can plot only a subset of its elements to take a closer look at its structure. With a moving ‘block’ from the beginning to the end of a vector or matrix or any R objects to which we can apply subset, all elements inside the block are plotted as a line or scatter plot or any customized plots.

For a vector, the elments from i + 1 to i + block will be plotted in the i-th step; similarly for a matrix or data frame, a (scatter) plot will be created from the i + 1-th row to i + block-th row.

However, this function is not limited to scatter plots or lines – we can customize the function FUN as we wish.

library(animation)
## (1) Brownian motion block length: 101 (i.e. 300-200+1)
ani.options(nmax = 200, 0.1)
# plot y = dat against x = i + 1:block customize xlab and
# ylab with 'i' and 'block' restrict ylim using the range of
# 'dat'
moving.block(dat = cumsum(rnorm(300)), FUN = function(..., dat = dat, 
  i = i, block = block) {
  plot(..., x = i + 1:block, xlab = sprintf("block length = %d", 
    block), ylim = range(dat), ylab = sprintf("x[%s:%s]", 
    i + 1, i + block))
}, type = "o", pch = 20)
## (2) Word counts of Hu's speech (block = 10;
## length(HuSpeech) = 75) see any pattern in the President's
## speech?
ani.options(nmax = 66, 1)
moving.block(dat = HuSpeech, FUN = function(..., dat = dat, i = i, 
  block = block) {
  plot(..., x = i + 1:block, xlab = "paragraph index", ylim = range(dat), 
    ylab = sprintf("HuSpeech[%s:%s]", i + 1, i + block))
}, type = "o", pch = 20)
## (3) sunspot data: observe the 11-year cycles block = 11
## years x 12 months/year = 132 set interval greater than 0 if
## your computer really rocks!
ani.options(nmax = 2857, 0.1)
spt.att = tsp(sunspot.month)
# the time index (we need it to correctly draw the ticks of
# x-axis)
ts.idx = seq(spt.att[1], spt.att[2], 1/spt.att[3])
moving.block(dat = sunspot.month, block = 132, FUN = function(..., 
  dat = dat, i = i, block = block) {
  plot(..., x = ts.idx[i + 1:block], xlab = sprintf("block length = %d", 
    block), ylim = range(dat), ylab = sprintf("sunspot.month[%s:%s]", 
    i + 1, i + block))
}, type = "o", pch = 20)
## Warning in moving.block(dat = sunspot.month, block = 132,
## FUN = function(..., : block length is too short; try to
## adjust 'block' or ani.options('nmax')
## (4) earth quake: order the data by 'depth' first see how
## the locations change as 'depth' increases
ani.options(nmax = 900, 0.1)
# compute the mean depth for each block of data
moving.block(quakes[order(quakes$depth), c("long", "lat")], FUN = function(..., 
  dat = dat, i = i, block = block) {
  plot(..., xlab = sprintf("%s[%s:%s]", colnames(dat)[1], i + 
    1, i + block), ylab = sprintf("%s[%s:%s]", colnames(dat)[2], 
    i + 1, i + block), xlim = range(dat[, 1]), ylim = range(dat[, 
    2]), main = sprintf("Mean Depth = %.3f", mean(sort(quakes$depth)[i + 
    1:block])))
}, pch = 20, col = rgb(0, 0, 0, 0.5))