cd_plot function

Conditional Density Plots

Computes and plots conditional densities describing how the distribution of a categorical variable y changes over a numerical variable x.

cd_plot(x, ...)
## Default S3 method:
cd_plot(x, y,
  plot = TRUE, ylab_tol = 0.05,
  bw = "nrd0", n = 512, from = NULL, to = NULL,
  main = "", xlab = NULL, ylab = NULL, margins = c(5.1, 4.1, 4.1, 3.1),
  gp = gpar(), name = "cd_plot", newpage = TRUE, pop = TRUE, return_grob = FALSE, ...)
## S3 method for class 'formula'
cd_plot(formula, data = list(),
  plot = TRUE, ylab_tol = 0.05,
  bw = "nrd0", n = 512, from = NULL, to = NULL,
  main = "", xlab = NULL, ylab = NULL, margins = c(5.1, 4.1, 4.1, 3.1),
  gp = gpar(), name = "cd_plot", newpage = TRUE, pop = TRUE, return_grob = FALSE, ...)

Arguments

• x: an object, the default method expects either a single numerical variable.

• y: a "factor" interpreted to be the dependent variable

• formula: a "formula" of type y ~ x with a single dependent "factor"

  and a single numerical explanatory variable.

• data: an optional data frame.

• plot: logical. Should the computed conditional densities be plotted?

• ylab_tol: convenience tolerance parameter for y-axis annotation. If the distance between two labels drops under this threshold, they are plotted equidistantly.

• bw, n, from, to, ...``: arguments passed to density

• main, xlab, ylab: character strings for annotation

• margins: margins when calling plotViewport

• gp: a "gpar" object controlling the grid graphical parameters of the rectangles. It should specify in particular a vector of fill colors of the same length as levels(y). The default is to call gray.colors.

• name: name of the plotting viewport.

• newpage: logical. Should grid.newpage be called before plotting?

• return_grob: logical. Should a snapshot of the display be returned as a grid grob?

• pop: logical. Should the viewport created be popped?

Details

cd_plot computes the conditional densities of x given the levels of y weighted by the marginal distribution of y. The densities are derived cumulatively over the levels of y.

This visualization technique is similar to spinograms (see spine) but they do not discretize the explanatory variable, but rather use a smoothing approach. Furthermore, the original x axis and not a distorted x axis (as for spinograms) is used. This typically results in conditional densities that are based on very few observations in the margins: hence, the estimates are less reliable there.

Returns

The conditional density functions (cumulative over the levels of y) are returned invisibly.

See Also

spine, density

References

Hofmann, H., Theus, M. (2005), Interactive graphics for visualizing conditional distributions, Unpublished Manuscript.

Author(s)

Achim Zeileis Achim.Zeileis@R-project.org

Examples

## Arthritis data
data("Arthritis")
cd_plot(Improved ~ Age, data = Arthritis)
cd_plot(Improved ~ Age, data = Arthritis, bw = 3)
cd_plot(Improved ~ Age, data = Arthritis, bw = "SJ")
## compare with spinogram
spine(Improved ~ Age, data = Arthritis, breaks = 3)

## Space shuttle data
data("SpaceShuttle")
cd_plot(Fail ~ Temperature, data = SpaceShuttle, bw = 2)

## scatter plot with conditional density
cdens <- cd_plot(Fail ~ Temperature, data = SpaceShuttle, bw = 2, plot = FALSE)
plot(I(-1 * (as.numeric(Fail) - 2)) ~ jitter(Temperature, factor = 2), data = SpaceShuttle,
  xlab = "Temperature", ylab = "Failure")
lines(53:81, cdens[[1]](53:81), col = 2)