testFraction {SafeVote}R Documentation

Bootstrapping experiment, with fractional counts of a ballot box.

Description

Starting from some number (astart) of randomly-selected ballots, an increasingly-large collection of randomly-selected ballots are counted. The ballots are chosen independently without replacement for each experimental unit; if you want to count decreasingly-sized portions of a single sample of ballots, use testDeletions().

Usage

testFraction(
  votes = NULL,
  astart = NULL,
  ainc = NULL,
  arep = NULL,
  trep = NULL,
  rankMethod = "safeRank",
  countMethod = "stv",
  countArgs = list(),
  exptName = NULL,
  equiet = FALSE,
  everbose = FALSE
)

Arguments

votes

A numeric matrix: one row per ballot, one column per candidate

astart

Starting number of ballots (min 2)

ainc

Number of ballots to be added in each step. Must be non-negative.

arep

Number of repetitions of the test on each step. Required to be non-NULL if ainc=0 && is.null(trep)'.

trep

Limit on the total number of simulated elections. Required to be non-NULL if ainc=0 && is.null(arep).

rankMethod

"safeRank" (default), "elected", or "rank". "rank" is a total ranking of the candidates, with ties broken at random. "elected" assigns rank=1 to elected candidates, rank=2 for eliminated candidates.

countMethod

countMethod "stv" (default) or "condorcet"

countArgs

List of args to be passed to countMethod (in addition to votes)

exptName

stem-name of experimental units e.g. "E". If NULL, then a 3-character string of capital letters is chosen at random.

equiet

TRUE to suppress all experimental output

everbose

TRUE to produce diagnostic output from the experiment

Value

a SafeRankExpt object of experimental results.

Examples

data(food_election)
testFraction(food_election, countMethod="condorcet",
             countArgs=list(safety=0.5,complete.ranking=TRUE))
testFraction(dublin_west, astart=20, ainc=10, arep=2, trep=3, 
             countMethod="stv", rankMethod="elected", equiet=FALSE)

[Package SafeVote version 1.0.0 Index]