ShapeChange (version 1.5)

jp: Specify a Jump-Point Category in a CHANGEPT Formula

Description

A symbolic routine to define that the underlying mean curve has a jump point in the formula argument to changept.

Usage

jp(x, trd1 = -1, trd2 = -1, up = TRUE)

Value

The vector "x" with five attributes, i.e., nm: the name of x; categ: the category of the change-point, "jump"; trd1 and trd2: \(-1\)(non-increasing) or \(1\) (non-decreasing); up: the jump direction.

The nm and categ attributes are used in the plot routine; the trd1, trd2, and up attributes are used to set up a shape-constrained regression.

Arguments

x

The predictor variable.

trd1

If trd1 is \(-1\) (\(1\)), then the estimated curve is constrained to be non-increasing (non-decreasing) on the left of the estimated jump point. The default is trd1 = \(-1\).

trd2

If trd2 is \(-1\) (\(1\)), then the estimated curve is constrained to be non-increasing (non-decreasing) on the right of the estimated jump point. The default is trd2 = \(-1\).

up

The jump direction at the jump point. The default is up = TRUE, i.e., there is an upward jump; otherwise, there is a downward jump.

Author

Xiyue Liao

Details

For the jump-point case, the user can choose either to impose a shape constraint on the estimated curve or not. If no shape constraint is imposed, only a jump point will be detected by the changept routine; otherwise, a jump point will be detected and the shape of the curve will be constrained according to the parameters trd1 and trd2 in a subinterval which is made of two consecutive knots and contains the jump point.

See Also

tp, ip

Examples

if (FALSE) {
  # simulated curve with an upward jump at .7
  n = 1000
  x = seq(1/n, 1, length = n)

  set.seed(123) 
  y =  4 * sin(5 * x) + 3 * x + (x >= .7) + rnorm(n, sd = 1)
  ans = changept(y ~ jp(x))

  plot(ans)
}