Plotting paths to data enlightenment
The package zenplots (zigzag expanded navigation plots) displays high-dimensional data by laying out alternating one- and two-dimensional plots in a zigzag-like pattern where adjacent axes share the same variate.
Zenplots are especially useful when subsets of pairs can be identified as of particular interest by some measure, or as not meaningfully comparable, or when pairs of variates can be ordered in terms of potential interest to view, or the number of pairs is too large for more traditional layouts such as a scatterplot matrix. They also allow an essentially arbitrary layout of plots.
High dimensional space can be explored in a zenplot
(zenplot()) by navigating through lower dimensional
subspaces along a zenpath (zenpath()) which orders the
dimensions (i.e. variates) visited according to some measure of
interestingness; see Hofert
and Oldford (2018)) for an application to S&P 500 constituent
data or Hofert
et al (2019) for application to visualizing dependence between two
sets of variates. See Hofert and Oldford
(2020) in Journal of
Statistical Software for algorithms.
The R package zenplots provides compact displays for
high-dimensional data via the notion of zenplots, grouping of variates,
and customizable displays of zigzag layouts. It accommodates different
graphical systems including the base graphics package, the
grid package (and hence packages like
ggplot2), and the interactive graphical package
loon.
zenplots handles groups of variates, partial and fully
missing data, and more.
One important feature is that zenplot() and its
auxiliary functions in zenplots distinguish layout from plotting which
allows one to freely choose and create one- and two-dimensional plot
functions; predefined functions are exported for all graphical
systems.
Try it.
Become one with your data!