guided_pcp: Guided parallel coordinate plot.

Description

Draws a parallel coordinate plot, with an accompanying barchart showing an index (eg correlation, scagnostics) levels for each panel. An index legend is optional.

Usage

guided_pcp(data, edgew=NULL, path = NULL, pathw=NULL,zoom=NULL,pcpfn=pcp,
     pcp.col = 1,lwd=0.5, panel.colors=NULL, pcp.mar=c(1.5,2,2,2), pcp.scale=TRUE,
     bar.col=1:9,bar.axes=FALSE, bar.mar=NULL,bar.ylim=NULL, reorder.weights=TRUE,
    layout.heights=NULL, layout.widths=c(10,1),
     main=NULL,legend=FALSE,cex.legend = 1,legend.mar=c(1,4,1,1),...)

Arguments

data: A data frame or matrix.
edgew: Matrix (or vector) whose rows give index values for each pair of variables.
path: an index vector specifying variable order, or a function. If a function, find_path(edgew,path,...) constructs the index vector.
pathw: Matrix (or vector) whose rows give index values for each adjacent pair of variables in path. Usually this argument is NULL and pathw is computed from the path and edgew.
zoom: If provided, a numeric vector specifying a subsequence of path to display.
pcpfn: Function to draw the parallel coordinates.
pcp.col: Line colors.
lwd: Line widths.
panel.colors: Background panel colors, passed to thepcpfn
pcp.mar: Controls PCP margin size.
pcp.scale: If TRUE, the variables will be scaled to 0-1 range, otherwise the data is not scaled.
bar.col: Bar colors.
bar.axes: Draw barplot axes, if TRUE.
bar.mar: Controls barplot margin size.
bar.ylim: Vertical limits of bar plot.
reorder.weights: If TRUE, reorder barplot indices so large values are drawn at the bottom.
layout.heights: Controls the layout.
layout.widths: Controls the layout.
main: Main title for PCP.
legend: If TRUE, draws the barplot index legend.
cex.legend: Controls legend text size.
legend.mar: Legend margin size.
...: Optional arguments

Author

C.B. Hurley and R.W. Oldford

References

see overview

Examples

Run this code

	 			
require(PairViz) 			


data <- mtcars[,c(1,3:6)]
cols <- c("red","green")[mtcars[,9]+1 ]    # transmission type, red=automatic

# add a correlation guide and find "better" hamiltonians...

# add a correlation guide...


corw <- dist2edge(as.dist(cor(data)))
edgew <- cbind(corw*(corw>0), corw*(corw<0))
         
# add a correlation guide to a PCP, positive cors shown in blue, negative in purple...

if (FALSE) {
dev.new(width=3,height=3)

par(cex.axis=.65)

guided_pcp(data,edgew, pcp.col=cols,
         main="Correlation guided PCP",bar.col = c("blue","purple"))

dev.new(width=7,height=3)
par(cex.axis=.65)

guided_pcp(data,edgew, path=eulerian, pcp.col=cols,lwd=2,
         main="Correlation guided Eulerian PCP",bar.col = c("blue","purple"),bar.axes=TRUE)



}

# Scagnostic guides are useful here- see the demos for more examples.

Run the code above in your browser using DataLab

Data engineering and BI courses are free!