SuperRanker (version 1.2.1)

random_list_sra: Simulate sequential rank agreement for randomized unrelated lists

Description

Simulate sequential rank agreement from completely uninformative lists (ie., raw permutations of items) and compute the corresponding sequential rank agreement curves. The following attributes are copied from the input object: number of lists, number of items and amount of censoring.

Usage

random_list_sra(
  object,
  B = 1,
  n = 1,
  na.strings = NULL,
  nitems = NULL,
  type = c("sd", "mad"),
  epsilon = 0
)

Value

A matrix with n columns and the same number of rows as for the input object. Each column contains one simulated sequential rank agreement curve from one permutation run.

Arguments

object

A matrix of numbers or list of vectors representing ranked lists.

B

An integer giving the number of randomizations to sample over in the case of censored observations

n

Integer: the number of permutation runs. For each permutation run we permute each of the lists in object and compute corresponding the sequential rank agreement curves

na.strings

A vector of character values that represent censored observations

nitems

The total number of items in the original lists if we only have partial lists available. Will be derived from the unique elements of the object if set to NULL (the default)

type

The type of measure to use. Either sd (standard deviation - the default) or mad (median absolute deviance)

epsilon

A non-negative numeric vector that contains the minimum limit in proportion of lists that must show the item. Defaults to 0. If a single number is provided then the value will be recycles to the number of items. Should usually be low.

Author

Claus Ekstrøm <[e-mail]>

Examples

# setting with 3 lists
mlist <- matrix(cbind(1:8,c(1,2,3,5,6,7,4,8),c(1,5,3,4,2,8,7,6)),ncol=3)
# compute sequential rank agreement of lists
sra(mlist)
# compute sequential rank agreement of 5 random permutations
random_list_sra(mlist, n=5)