ror (version 1.2)

sample.vfs.rejection: Robust Ordinal Regression Value Function sampler

Description

Pure rejection sampling of general monotone value functions. The current version assumes ascending preferences, i.e. higher criterion evaluation means higher preferability (=better).

Usage

sample.vfs.rejection(performances, preferences, nr=10000, updInterval=1000)

Arguments

performances

m x n performance matrix with m alternatives and n criteria

preferences

k x 2 matrix of preferences statements (row indices of alternatives in performance matrix). Each row r is a preference statements meaning that preferences[k,1] is preferred to preferences[k,2]

nr

The number of value functions to sample

updInterval

Update interval for printing out current iteration. 0 = no information will be printed out

Value

value functions for the i'th criterion, one function per row and 'misses' contains the amount of misses during value function sampling.

See Also

rorsmaa,utagms,ror-package,sample.vfs.gibbs

Examples

# Set Java VM memory use to 2g not to run out of heap space
options( java.parameters = "-Xmx2g" )
library(ror)

## Function needed to generate pareto-optimal alternatives
randomPointFromHypersphere <- function(ncrit) {
  rns <- c()
  while(TRUE) {
    rns <- rnorm(ncrit)
    if (all(rns > 0)) {
      break
    }
  }
  mul <- 1 / sqrt(sum(rns * rns))
  return(rns * mul)
}

performances <- t(replicate(10, randomPointFromHypersphere(5)))  # 10 alts, 5 crit
preferences <- matrix(c(1, 2, 4, 5, 7, 8, 1, 3), ncol=2, byrow=TRUE)

vfs <- sample.vfs.rejection(performances, preferences, nr=10)