rsm (version 1.30)

contour.lm: Surface plot(s) of a fitted linear model

Description

contour and persp methods that will work with any lm object involving two or more numerical predictors.

Usage

## S3 method for class 'lm':
contour(x, form, at, bounds, zlim, xlabs, hook, 
  plot.it=TRUE, image=FALSE, img.col=terrain.colors(50), ...)

## S3 method for class 'lm':
image(x, form, at, bounds, zlim, xlabs, hook,  ...)

## S3 method for class 'lm':
persp(x, form, at, bounds, zlim, zlab, xlabs, col = "white", 
  contours = NULL, hook, theta = -25, phi = 20, r = 4, 
  border = NULL, box = TRUE, ticktype = "detailed", ...)

Arguments

x

A lm object.

form

A formula, or a list of formulas.

at

Optional named list of fixed values to use for variables other than the axis variables. If not provided, the mean is used for numeric variables, or the first level if it is a factor.

bounds

Optional named list of bounds or grid values to use for the variables having the same names. See details.

zlim

Optional zlim setting passed to image. If not provided, the range of values across all plotted surfaces is used.

zlab

Optional label for the vertical axis.

xlabs

Alternate labels for predictor axes (see Details).

hook

Optional list that can contain functions pre.plot and post.plot. May be used to add annotations or to re-route the graphs to separate files (see Details).

image

Set to TRUE if you want an image plot overlaid by contours.

img.col

Color map to use when image=TRUE.

plot.it

If TRUE, no plot is produced, just the return value.

col

Color or colors used for facets in the perspective plot (see details).

contours

If non-NULL, specifications for added contour lines in perspective plot.

theta, phi

Viewing angles passed to persp (different defaults).

r

Viewing distance passed to persp (different default).

border, box

Options passed to persp.

ticktype

Option passed to persp (different default).

...

Additional arguments passed to contour or persp.

Value

• A list containing information that is plotted. Each list item is itself a list with the following components:
• x, y
  The values used for the x and y axes
• z
  The matrix of fitted response values
• labs
  Character vector of length 4: Elements 1 and 2 are the x and y axis labels, and elements 3 and 4 are their original variable names
• zlim
  The computed or provided zlim values
• transf
  (persp only) The 3D transformation for trans3d

Details

form may be a single formula or a list of formulas. A simple formula like x2 ~ x1 will produce a contour plot of the fitted regression surface for combinations of x2 (vertical axis) and x1 (horizontal axis). A list of several such simple formulas will produce a contour plot for each formula. A two-sided formula produces contour plots for each left-hand variable versus each right-hand variable (except when they are the same); for example, x1+x3 ~ x2+x3 is equivalent to list(x1~x2, x3~x2, x1~x3). A one-sided formula produces contour plots for each pair of variables. For example, ~ x1+x2+x3 is equivalent to list(x2~x1, x3~x1, x3~x2). For any variables not in the bounds argument, a grid of 26 equally-spaced values in the observed range of that variable is used. If you specify a vector of length 2, it is interpreted as the desired range for that variable and a grid of 26 equally-spaced points is generated. If it is a vector of length 3, the first two elements are used as the range, and the third as the number of grid points. If it is a vector of length 4 or more, those values are used directly as the grid values. By default, the predictor axes are labeled using the variable names in form, unless x is an rsm object, in which case the variable-coding formulas, if present, are used to generate axis labels. These labels are replaced by the entries in xlabs if provided. One must be careful using this to make sure that the names are mapped correctly. The entries in xlabs should match the respective unique variable names in form, after sorting them in (case-insensitive) alphabetical order. Note that if form is changed, it may also be necessary to change xlabs. In persp, contour lines may be added via the contours argument. It may be a boolean or character value, or a list. If boolean and TRUE, default black contour lines are added to the bottom surface of the box. Character values of "top", "bottom" add black contour lines to the specified surface of the box. contours = "colors" puts contour lines on the bottom using the same colors as those at the same height on the surface. Other character values of contours are taken to be the desired color of the contour lines, plotted at the bottom. If contours is a list, its elements (all are optional) are used as follows: [object Object],[object Object],[object Object] Since these functions often produce several plots, the hook argument is provided if special setups or annotations are needed for each plot. It should be a list that defines one or both of the functions pre.plot and post.plot. Both of these functions have one argument, the character vector labs for that plot (see Value documentation). Additional examples and discussion of these plotting functions is available via vignette("rsm-plots").

References

Lenth RV (2009). ``Response-Surface Methods in R, Using rsm'', Journal of Statistical Software, 32(7), 1--17. http://www.jstatsoft.org/v32/i07/.

See Also

contour

Examples

library (rsm)
heli.rsm = rsm (ave ~ block + SO(x1, x2, x3, x4), data = heli)

# Plain contour plots
par (mfrow = c (2,3))
contour (heli.rsm, ~x1+x2+x3+x4, at=canonical(heli.rsm)$xs)

# Same with image overlay and extra annotations showing 
# stationary point and "at" values
# Demonstrates use of 'hook' argument
xs = canonical(heli.rsm)$xs
myhook = list()
myhook$post.plot = function(lab) {
  idx = sapply(lab[3:4], grep, names(xs))
  points (xs[idx[1]], xs[idx[2]], pch=2, col="red")
  main = paste("Predicted response versus", lab[3], "and", lab[4])  
  atlabs = paste(names(xs[-idx]), round(xs[-idx], 3), sep = "= ")
  atlabs = paste(atlabs, collapse = ", ")
  title(c(main, paste("at", atlabs)))
}
contour (heli.rsm, ~x1+x2+x3+x4, at=xs, hook=myhook, image=TRUE)

# Default perspective views
persp (heli.rsm, ~x1+x2+x3+x4, at=canonical(heli.rsm)$xs)

# Same plots, souped-up with facet coloring and axis labeling
persp (heli.rsm, ~x1+x2+x3+x4, contours="col", col=rainbow(40), at=canonical(heli.rsm)$xs,
  xlabs = c("Wing area", "Wing length", "Body width", "Body length"), zlab = "Flight time")
  
### Hints for creating graphics files for use in publications...

# Save perspective plots in one PDF file (will be six pages long)
pdf(file = "heli-plots.pdf")
persp (heli.rsm, ~x1+x2+x3+x4, at=canonical(heli.rsm)$xs)
dev.off()

# Save perspective plots in six separate PNG files
png.hook = list()
png.hook$pre.plot = function(lab) 
    png(file = paste(lab[3], lab[4], ".png", sep = ""))
png.hook$post.plot = function(lab)
    dev.off()
persp (heli.rsm, ~x1+x2+x3+x4, at=canonical(heli.rsm)$xs, hook = png.hook)