Conditioning Plot

DESCRIPTION:

The coplot function produces a Conditioning Plot in the current graphics device which shows how a response depends on a predictor given another predictor.

USAGE:

coplot(formula, data, given.values, panel = points, rows, columns, 
     column.row, show.given = T, add = F, xlab, ylab, xlim, ylim, ...)

REQUIRED ARGUMENTS:

formula: formula defining the response and the predictors involved in the plotting.

OPTIONAL ARGUMENTS:

data: data frame in which the formula will be evaluated. If missing, evaluation will take place as if the formula were evaluated in the frame of the function calling coplot.
given.values: a numeric vector, character vector, or two-column matrix that specifies the given values when there is one given predictor, or a list of two such objects when there are two. If missing, reasonable things happen.
panel: a user-supplied function of x and y that determines the method of plotting on the dependence panels.
rows: for the case of one given predictor, the number of rows of the matrix of dependence panels. If missing, the following is the default: let k be the number of given values; if columns is missing, then it is ceiling(sqrt(k)) and otherwise it is ceiling(k/columns). This argument is not used if there are two given predictors.
columns: for the case of one given predictor, the number of columns of the matrix of dependence panels. If missing, the following is the default: let k be the number of given values; if rows is missing, it is ceiling(k/ceiling(sqrt(k))) and otherwise it is ceiling(k/rows). This argument is not used if there are two given predictors.
column.row: when conditioning on only one predictor, a vector of length 2 describing the arrangement of panels in the plot, i.e. a vector of the form c(n,m) where n and m are integers indicating the number of columns and rows, respectively for the arrangement of the dependence panels. The default is given by the arguments columns and rows which are overriden by this argument, if specified.
show.given: if FALSE, given panels are not included.
add: if TRUE, add to the current plot. Graphical parameters may also be supplied as arguments to this function (see par ). The arguments xlab and ylab are as for other graphics functions, except that the former is a character vector of labels for predictors. If the elements have names, they are matched to the names in formula; if not, the elements are assigned, in order, according to the order in which they appear in formula. If missing, the names of the predictors in formula are used. The arguments xlim and ylim are axis limits as in other graphics functions.

SIDE EFFECTS:

a plot is produced on the current graphics device. par(mfrow) is reset to c(1,1) before the plot is drawn, that is, the plot will take up the entire graphics area; no other plot can appear at the same time.

BUGS:

coplot does not recognize graphics parameters if the panel argument is suppled as a function definition (i.e., panel=function(x,y)...). In this case, you must supply the graphics parameters in the definition of the panel function.

EXAMPLES:

# the following makes a coplot of NOx against C given E 
# with smoothings of the scatterplots on the dependence panels: 
E.intervals <- co.intervals(ethanol$E, 16, 0.25) 
coplot(NOx ~ C | E, given.values = E.intervals, data = ethanol, 
      panel = function(x, y) panel.smooth(x, y, span = 1, 
      degree = 1))