Last chance! 50% off unlimited learning
Sale ends in
This plot can be used to graphically display a correlation matrix by using the linewidth between the nodes in proportion to the correlation of two variables. It will place the elements homogenously around a circle and draw connecting lines between the points.
PlotWeb(m, col = c(hred, hblue), lty = NULL, lwd = NULL, args.legend=NULL,
pch = 21, pt.cex = 2, pt.col = "black", pt.bg = "darkgrey",
cex.lab = 1, las = 1, adj = NULL, dist = 0.5, ...)a symmetric matrix of numeric values
the color for the connecting lines
the line type for the connecting lines, the default will be par("lty").
the line widths for the connecting lines. If left to NULL it will be linearly scaled between the minimum and maximum value of m.
list of additional arguments to be passed to the legend function.
Use args.legend = NA if no legend should be added.
the plotting symbols appearing in the plot, as a non-negative numeric vector (see points, but unlike there negative values are omitted) or a vector of 1-character strings, or one multi-character string.
expansion factor(s) for the points.
the foreground color for the points, corresponding to its argument col.
the background color for the points, corresponding to its argument bg.
alignment of the labels, 1 means horizontal, 2 radial and 3 vertical.
adjustments for the labels. (Left: 0, Right: 1, Mid: 0.5)
gives the distance of the labels from the outer circle. Default is 2.
the character extension for the labels.
dots are passed to plot.
A list of x and y coordinates, giving the coordinates of all the points drawn, useful for adding other elements to the plot.
The function uses the lower triangular matrix of m, so this is the order colors, linewidth etc. must be given, when the defaults are to be overrun.
# NOT RUN {
m <- cor(d.pizza[, which(sapply(d.pizza, IsNumeric, na.rm=TRUE))[-c(1:2)]],
use="pairwise.complete.obs")
PlotWeb(m=m, col=c(hred, hblue), main="Pizza Correlation")
# let's describe only the significant corrs and start with a dataset
d.m <- d.pizza[, which(sapply(d.pizza, IsNumeric, na.rm=TRUE))[-c(1:2)]]
# get the correlation matrix
m <- cor(d.m, use="pairwise.complete.obs")
# let's get rid of all non significant correlations
ctest <- PairApply(d.m, function(x, y) cor.test(x, y)$p.value, symmetric=TRUE)
# ok, got all the p-values, now replace > 0.05 with NAs
m[ctest > 0.05] <- NA
# How does that look like now?
Format(m, na.form = ". ", leading="drop", digits=3, align = "right")
PlotWeb(m, las=2, cex=1.2)
# define line widths
PlotWeb(m, lwd=abs(m[lower.tri(m)] * 10))
# }
Run the code above in your browser using DataLab