Exploratory graphics

Basic plots

Single variable:

• histogram/density plot
• qqplots

Two variables:

• scatterplot (may need multiple plot to distinguish different groups)
• boxplots

More than two variables:

• cloud plot (brush & spin etc.)
• coplots (show relationship between x and y, conditioned on z)

Trellis graphics

Can generalise the coplot to condition on more than one variable. Easy if the variables are categorical, but if they’re continuous we need to divide them up in to subranges. Conditoning variables determine layout of cells. x/y variables determine type of graph in each cell.

In R:

library(lattice)
xyplot(y ~ x | w * z, data=a.df)
xyplot(y ~ x | equal.count(w, number=4, overlap=0)