unscaleDesign {DiceDesign}

Unscale a Design

Description

This function unscales the values of a scaled design (values in [0,1]). The unscaling can be made by the inverse Rosenblatt transformation (by applying the empirical quantile function given by another design) or by translating the design from maximum and minimum values (given for each variable).

Usage

unscaleDesign(design, min=NULL, max=NULL, uniformize=FALSE, InitialDesign=NULL)

Arguments

design

a matrix (or a data.frame) corresponding to the design of experiments to unscale

min

the vector of minimal bounds of each design variable

max

the vector of maximal bounds of each design variable

uniformize

boolean: TRUE to use the inverse Rosenblatt transformation (the min and max vectors are useless in this case). If FALSE (default value), the translation from max and min values is applied

InitialDesign

If the inverse Rosenblatt transformation is applied (uniformize = TRUE): a matrix (or a data.frame) corresponding to the design which gives the empirical quantiles

Value

A list containing:

design

the unscaled design

min

the vector of minimal bounds that has been used

max

the vector of maximal bounds that has been used

uniformize

the value of this boolean argument

Author(s)

B. Iooss

Examples

  d <- 2
  n <- 100
  x <- matrix(rnorm(d*n), ncol=d)
  xscale <- scaleDesign(x, uniformize=TRUE)
  xunscale1 <- unscaleDesign(xscale$design, uniformize=TRUE, InitialDesign=x)
  xunscale2 <- unscaleDesign(xscale$design, 
  min=c(min(x[,1]), min(x[,2])), max = c(max(x[,1]), max(x[,2])))
  par(mfrow=c(2,2))
  plot(x) ; plot(xscale$design)
  plot(xunscale1$design) ; plot(xunscale2$design)