Lattice Graphics

Description

Implementation of Trellis Graphics in R

Details

Trellis Graphics is a framework for data visualization developed at the Bell Labs by Rick Becker, Bill Cleveland et al, extending ideas presented in Bill Cleveland's 1993 book Visualizing Data.

Lattice is best thought of as an implementation of Trellis Graphics for R. Its interface is based on the implementation in S-PLUS, but there are several differences. To the extent possible, care has been taken to ensure that existing Trellis code written for S-PLUS works unchanged (or with minimal change) in Lattice. If you are having problems porting S-PLUS code, read the entry for panel in the documentation for xyplot. Most high level Trellis functions in S-PLUS are implemented, with the exception of piechart.

Lattice is built upon the Grid Graphics engine for R being developed by Paul Murrell and requires the grid add-on package.

Type help(package = lattice) to see a list of (public) Lattice graphics functions for which further documentation is available. The ‘See Also’ section below has list of specific areas of possible interest and pointers to the help pages with respective details. Apart from the documentation accompanying this package, several documents outlining the use of Trellis graphics is available from Bell Lab's website that might provide a holistic introduction to the Trellis paradigm.

Note

High level Lattice functions (like xyplot) are different from conventional R graphics functions because they don't actually draw anything. Instead, they return an object of class "trellis" which has to be then printed or plotted to create the actual plot. This is normally done automatically, but not when the high level functions are called inside another function (most often source) or other contexts where automatic printing is suppressed (e.g. for or while loops). In such situations, an explicit call to print or plot is required.

Lattice plots are highly customizable via user-modifiable settings. However, these are completely unrelated to base graphics settings; in particular, changing par() settings usually have no effect on lattice plots.

Author(s)

Deepayan Sarkar Deepayan.Sarkar@R-project.org

References

Bell Lab's Trellis Page: http://cm.bell-labs.com/cm/ms/departments/sia/project/trellis/

Cleveland, W.S. (1993) Visualizing Data.

Becker, R.A., Cleveland, W.S. and Shyu, M. “The Visual Design and Control of Trellis Display”, Journal of Computational and Graphical Statistics

See Also

The Lattice user interface primarily consists of several ‘high level’ generic functions (listed below), each designed to create a particular type of statistical display by default. While each function does different things, they share several common features, reflected in several common arguments that affect the resulting displays in similar ways. These arguments are extensively (sometimes only) documented in the help page for xyplot. This includes a discussion of conditioning and control of the Trellis layout.

Lattice employs an extensive system of user-controllable parameters to determine the look and feel of the displays it produces. To learn how to use and customise the Graphical parameters used by the Lattice functions, see trellis.par.set. For other settings, see lattice.options. The default graphical settings are different for different graphical devices. To learn how to initialise new devices with the desired settings or change the settings of the current device, see trellis.device.

To learn about sophisticated (non-default) printing capabilities, see print.trellis. See update.trellis to learn about manipulating a "trellis" object. Tools to augment lattice plots after they are drawn (including locator-like functionality) is described in the interaction help page.

The following is a list of ‘high level’ functions in the Lattice package with a brief description of what they do. In all cases, the actual display is produced by the so-called panel function, which has a suitable default, but can be substituted by an user defined function to create custom displays. The user will most often be interested in the default panel functions, which have a separate help page, linked to from the help pages of the corresponding high level function. Although documented separately, arguments to these panel functions can be supplied directly to the high level functions, which will forward the arguments as appropriate.

Univariate:

barchart bar plots

bwplot box and whisker plots

densityplot kernel density plots

dotplot dot plots

histogram histograms

qqmath quantile plots against mathematical distributions

stripplot 1-dimensional scatterplot

Bivariate:

qq q-q plot for comparing two distributions

xyplot scatter plot (and possibly a lot more)

Trivariate:

levelplot level plots (similar to image plots in R)

contourplot contour plots

cloud 3-D scatter plots

wireframe 3-D surfaces (similar to persp plots in R)

Hypervariate:

splom scatterplot matrix

parallel parallel coordinate plots

Miscellaneous:

rfs residual and fitted value plot (also see oneway)

tmd Tukey Mean-Difference plot

Additionally, there are several panel functions that do little by themselves, but can be useful components of custom panel functions. These are documented in panel.functions. Lattice also has a collection of convenience functions that correspond to the base graphics primitives lines, points, etc. They are implemented using Grid graphics, but try to be as close to the base versions as possible in terms of their argument list. These functions have imaginative names like llines or panel.lines and are often useful when writing (or porting from S-PLUS code) nontrivial panel functions.