In this article, Hans Gilde exposes the clever use of a heatmap hidden in the Bioconductor library. In his example, he describes a way to show different ‘observations’ on subjects, with the concept of time.

Financial indices, like the S&P 500 or the Dow Jones indices, are mathematically some kind of measure of overall market performance, and their calculations (usually in real-time) might not translate very well the different price movements of the underlying index components.

Combining the different features of the quantmod package (as described here), one can come up with a time-series based heatmap of some financial indices, like the S&P 500 for example.

Below is an adaptation of Hans Gilde’s ideas to the S&P 500 since 2007. Note that some symbols were removed while preparing the data. Note also how highly correlated the different components are.


# get the list of symbols
l <- readHTMLTable('')[[1]]
l <- as.vector(l$Ticker)
l <- l[c(-59, -71, -80, -124, -141, -147, -275, -283, -292, -299, -309, -316, -360, -378, -381, -406, -439, -470, -471)]

getMonthlyReturns <- function(sym) {
    y <- to.monthly(getSymbols(sym, auto.assign=FALSE, from='2007-01-01'))

d <- unlist(llply(l, getMonthlyReturns, .progress="text"))
# bounds at -10% and +10% for visual clarity
d[d < -10] <- -10
d[d > 10] <- 10

heatmap_2(t(matrix(d, ncol=481)), col=brewer.pal(9, 'PuBu'), Rowv=NA, Colv=NA, do.dendro=c(FALSE,FALSE), scale='none', legend=2, main="S&P 500 since 2007 (monthly returns)")