srd: Draws scaled rectangle diagrams to represent categorical...
In srd: Draws Scaled Rectangle Diagrams

Description Usage Arguments Details Value Author(s) References Examples

Draws scaled rectangle diagrams to represent categorical data.

srd(data = stop("'data' must be specified"), weight = NA,
  labelcell = "frequency", new = TRUE, criterion = "logl",
  colour = "earthy", labelrectangle = "inside", y = NA, w = NA,
  cells = FALSE, motion = "none", thin = NA)

`data`	A binary data matrix of ones and zeroes (or TRUE and FALSE) with k columns, with 1 or TRUE representing presence of an attribute that represents a rectangle.
`weight`	A frequency count vector attaches a frequency weight to each row of `d`.
`labelcell`	Attaches labels to cells. One of "none", "frequency", "area", "error", or "residual". Use "frequency" for cell frequencies (default), "area" for areas, "error" for percentage error, "residual" for Pearson residuals
`new`	TRUE or FALSE. Fits a new configuration, new=FALSE assumes an existing configuration already exists
`criterion`	Criterion to determine fitting optimisation. One of "lsqs", "labs", "logl", or "chi2". Use "lsqs" for least squares (default), "labs" for least absolute difference, "logl" for minus log-likelihood, "chi2" for chi-square
`colour`	Determines colour of configuration. One of "earthy", "bright", or "mono". Use "earthy" for muted colours, "bright" for rainbow colours (default), "mono" is monochrome (shades of grey). If `cells=TRUE`, "earthy" and "bright" are interpreted as shades of yellow and blue respectively.
`labelrectangle`	To attach labels to each of the k rectangles. One of "none", "inside", or "arrow". Use "none" for no labels, "inside" (default) for inside rectangle, "arrow" is an arrow pointer from outside of the unit square.
`y`	A vector of data (with same number of rows as `d`) for the third axis in a 3D projection . Default NA is equivalent to a vector of zeroes. Projected values are mean values of `y`, or ∑ y / ∑ w if `w` is specified. Projected values are of each rectangle, or if `cells=TRUE` of each cell (see `cells` parameter)
`w`	A numeric vector associated with `y`. Statistics of `y` for cells/rectangles (see cells argument) represented by ∑ y / ∑ w. Default NA gives 3D projection as mean values of `y` (equivalent to `w=1`). `w` can be used, for example, to obtain 3D projection axis as a "incidence rate" if `y` is a binary indicator of an outcome and `w` is person-time.
`cells`	TRUE or FALSE. True for a diagram shaded by mean of y in each cell of intersecting rectangles and 3D projection showing each cell. Default (FALSE) is transparent represention of each of k rectangles, and 3D projection the mean of `y` of each rectangle.
`motion`	Determines whether and how configuration is put into 3D mode. One of values "none", "frames", "movie". "none" for a flat single image, "frames" for broken series of rotated images, "movie" for an animated rotating movie.
`thin`	Parameter to attempt to avoid thin and elongated rectangles. The `criterion` is penalised by multiplying by x^thin where is x is maximal thinness (length/breadth) of the q rectangles. Recommended value <=1. Default NA is no penalising, equivalent to thin=0.

A function to visualise k attributes as a Venn-like diagram with areas scaled, as best as possible, to be proportional to frequency. It is a representation is of a 2^q contingency table. Up to six attributes (rectangles) can be displayed. Fitting is done by optimising the congruence between cell area and cell frequency by the chosen criterion. A flat diagram can be thrown into an animated 3D projection, showing mean/rate values of the associated y with motion equal to "frames" or "movie".

Error message, or indication of successful completion.

Roger Marshall, <rj.marshall@auckland.ac.nz>, The University of Auckland, New Zealand

Marshall, RJ. Displaying categorical data relationships by scaled rectangle diagrams. Statistics in Medicine 20, 1077-1088, 2001. Marshall, RJ. Scaled rectangle diagrams can be used to visualise clinical and epidemiological data. Journal of Clinical Epidemiology 58, 974-981, 2005. Marshall, RJ. Determining and visualising at-risk groups in case-control data. Journal of Epidemiology and Biostatistics 6, 343-348, 2001

library(datasets)
data(Titanic)
x <- data.frame(Titanic)
x$women<-  ifelse(x$Sex=="Female",1,0)
x$child <- ifelse(x$Age=="Child",1,0)
x$first <- ifelse(x$Class=="1st",1,0)
x$survived <- ifelse(x$Survived=="yes",1,0)
d <- x[c("women","child","first")]
srd(data=d,y=x$survived,weight=x$Freq,motion="frames")


library(survival)
data(pbc)
## create srd and show projection axis as incidence rate (by specifying w as time)
pbc$male <- ifelse(pbc$sex=="m",1,0)
d  <- with(pbc,cbind(ascites,hepato,spiders,male))
pbc$event <- ifelse(pbc$status==2,1,0)
srd(d, y=pbc$event, w=pbc$time, colour="earthy",
   labelrectangle="arrow",cells=FALSE,motion="frames", criterion="chi2", thin=0.5)