Plot Proximity Profile for a profile.prx Object

S3 method for plotting proximity-weighted mean profiles from a boostmtree prediction object computed with proximity = TRUE. A randomly selected (or user-specified) subject is used as the reference case. Subjects whose proximity score to the reference case exceeds the cut quantile are treated as neighbours; their smoothed fitted trajectories (dashed lines) and a proximity-weighted average trajectory (solid black line) are overlaid on the reference subject's smoothed trajectory (solid blue line).

Usage

# S3 method for class 'profile.prx'
plot(x, col = NULL, rnd.case = NULL, cut = 0.95, restrictX = TRUE, ...)

Arguments

x

An object of class profile.prx — a boostmtree predict object that contains a proximity matrix (i.e., produced with proximity = TRUE).

col

Optional colour vector. Must be indexed by original subject index — colours are looked up as col[i] where i is the subject's row number in the proximity matrix. A safe choice is a vector of length nrow(x$proximity). Defaults to 1 (black) for every matched neighbour.

rnd.case

Integer index of the reference subject. If NULL (default) a subject is chosen at random.

cut

Quantile threshold (0–1) for defining neighbours based on proximity score. Defaults to 0.95.

restrictX

Logical. If TRUE (default) the x-axis is restricted to the observation times of the reference subject.

...

Additional arguments passed to plot() (currently unused).

Value

Invisibly returns the covariate matrix rows corresponding to the matched neighbours.

See also

predict.boostmtree