R: Plot a Correlation Matrix

PlotCorr {DescTools}

R Documentation

Plot a Correlation Matrix

Description

This function produces a graphical display of a correlation matrix. The cells of the matrix can be shaded or colored to show the correlation value.

Usage

PlotCorr(x, cols = colorRampPalette(c(Pal()[2], "white",
                                      Pal()[1]), space = "rgb")(20),
         breaks = seq(-1, 1, length = length(cols) + 1),
         border = "grey", lwd = 1,
         args.colorlegend = NULL, xaxt = par("xaxt"), yaxt = par("yaxt"),
         cex.axis = 0.8, las = 2, mar = c(3, 8, 8, 8), mincor = 0,
         main = "", ...)

Arguments

`x`	x is a correlation matrix to be visualized.
`cols`	the colors for shading the matrix. Uses the package's option `"col1"` and `"col2"` as default.
`breaks`	a set of breakpoints for the colours: must give one more breakpoint than colour. These are passed to `image()` function. If breaks is specified then the algorithm used follows `cut`, so intervals are closed on the right and open on the left except for the lowest interval.
`border`	color for borders. The default is `grey`. Set this argument to `NA` if borders should be omitted.
`lwd`	line width for borders. Default is 1.
`args.colorlegend`	list of arguments for the `ColorLegend`. Use `NA` if no color legend should be painted.
`xaxt`	parameter to define, whether to draw an x-axis, defaults to `"n"`.
`yaxt`	parameter to define, whether to draw an y-axis, defaults to `"n"`.
`cex.axis`	character extension for the axis labels.
`las`	the style of axis labels.
`mar`	sets the margins, defaults to mar = c(3, 8, 8, 8) as we need a bit more room on the right.
`mincor`	numeric value between 0 and 1, defining the smallest correlation that is to be displayed. If this is >0 then all correlations with a lower value are suppressed.
`main`	character, the main title.
`...`	the dots are passed to the function `image`, which produces the plot.

Value

no values returned.

Author(s)

Andri Signorell <andri@signorell.net>

Examples

m <- cor(d.pizza[,sapply(d.pizza, IsNumeric, na.rm=TRUE)], use="pairwise.complete.obs")

PlotCorr(m, cols=colorRampPalette(c("red", "black", "green"), space = "rgb")(20))
PlotCorr(m, cols=colorRampPalette(c("red", "black", "green"), space = "rgb")(20),
         args.colorlegend=NA)

m <- PairApply(d.diamonds[, sapply(d.diamonds, is.factor)], CramerV, symmetric=TRUE)
PlotCorr(m, cols = colorRampPalette(c("white", "steelblue"), space = "rgb")(20),
         breaks=seq(0, 1, length=21), border="black",
         args.colorlegend = list(labels=sprintf("%.1f", seq(0, 1, length = 11)), frame=TRUE)
)
title(main="Cramer's V", line=2)
text(x=rep(1:ncol(m),ncol(m)), y=rep(1:ncol(m),each=ncol(m)),
     label=sprintf("%0.2f", m[,ncol(m):1]), cex=0.8, xpd=TRUE)

# Spearman correlation on ordinal factors
csp <- cor(data.frame(lapply(d.diamonds[,c("carat", "clarity", "cut", "polish",
                      "symmetry", "price")], as.numeric)), method="spearman")
PlotCorr(csp)

# some more colors
m <- cor(mtcars)
PlotCorr(m, col=Pal("RedWhiteBlue1", 100), border="grey",
         args.colorlegend=list(labels=Format(seq(-1,1,.25), digits=2), frame="grey"))

# display only correlation with a value > 0.7
PlotCorr(m, mincor = 0.7)
x <- matrix(rep(1:ncol(m),each=ncol(m)), ncol=ncol(m))
y <- matrix(rep(ncol(m):1,ncol(m)), ncol=ncol(m))
txt <- Format(m, d=3, leading = "drop")
idx <- upper.tri(matrix(x, ncol=ncol(m)), diag=FALSE)

# place the text on the upper triagonal matrix
text(x=x[idx], y=y[idx], label=txt[idx], cex=0.8, xpd=TRUE)

# put similiar correlations together
m <- cor(mtcars)
idx <- order.dendrogram(as.dendrogram(
          hclust(dist(cor(mtcars)), method = "mcquitty")
       ))
PlotCorr(m[idx, idx])

[Package DescTools version 0.99.24 Index]