jump.prim: Peeling trajectory jump

In PierreMasselot/primr: Patient rule-induction method

Description

Identifies a jump in the peeling trajectory of object.

Usage

jump.prim(object, rel.support = TRUE)

Arguments

object	A prim object resulting from a call to peeling.
rel.support	Logical indicating if the trajectory difference should be relative to the support for finding the jump (default to TRUE).

Details

Computes the (relative) trajectory differences of object:

(yfun[k] - yfun[k - 1])/(support[k - 1] - support[k])

and returns its maximum value. The rationale is that the biggest jump in peeling trajectory gives a good cut-off point for the peeling algorithm. If rel.support = FALSE, the denominator is not used in the differences calculation.

Value

A list with elements:

trajectory.difference	Numeric vector of the computed (relative) differences.
npeel.opt	Integer giving the npeel value of the highest difference.
final.box	The extracted box corresponding to npeel.opt. See extract.box.

References

Masselot P., Chebana F., Campagna C., Lavigne E., Ouarda T.B.M.J., Gosselin P. On threshold identification for weather-health warning systems. Submitted.

Examples

   # A simple bump
   set.seed(12345)
   x <- matrix(runif(2000), ncol = 2, dimnames = list(NULL, c("x1", "x2")))
   y <- 2 * x[,1] + 5 * x[,2] + 10 * (x[,1] >= .8 & x[,2] >= .5) + 
     rnorm(1000)
   # Peeling with alpha = 0.05 and beta.stop = 0.05
   peel_res <- peeling(y, x, beta.stop = 0.05)
   # Automatically choose the best box
   chosen <- jump.prim(peel_res)

PierreMasselot/primr documentation built on Feb. 5, 2021, 7:33 p.m.