standalone_monte_carlo_methods: Faster Monte Carlo estimation of pivot probabilities

In aeggers/pivotprobs: Methods for estimating the probability of ties and other important election events.

Faster Monte Carlo estimation of pivot probabilities

Description

These methods take simulated election results and return a list of pivot events, each with a pivot probability (integral) and P matrix (P). They produce identical results to election_event_probs() (or nearly so) but they are faster because they are written in a less general way.

Usage

plurality_event_probs_from_sims(
  sims = NULL,
  n = 1000,
  window = 0.01,
  cand_names = NULL,
  sep = "_",
  method = "density",
  merge = F,
  drop = F,
  bw_divisor = 1,
  skip_non_pivot_events = T,
  s = NULL,
  raw = F
)

positional_event_probs_from_sims(
  sims,
  window = 0.01,
  n = 1000,
  s = 0.5,
  cand_names = NULL,
  sep = "_",
  method = "density",
  merge = F,
  drop = F,
  increments = 10,
  bw_divisor = 1,
  skip_non_pivot_events = T,
  raw = F
)

irv_event_probs_from_sims(
  sims,
  window = 0.01,
  n = 1000,
  s = 0,
  cand_names = NULL,
  sep = "_",
  method = "density",
  merge = F,
  bw_divisor = 1,
  skip_non_pivot_events = T
)

condorcet_event_probs_from_sims(
  sims,
  n = 1000,
  window = 0.01,
  cand_names = NULL,
  sep = "_",
  kemeny = T,
  method = "density",
  merge = F,
  bw_divisor = 1,
  s = NULL,
  skip_non_pivot_events = T
)

Arguments

sims	A matrix of simulated election results, with one column per ballot type. Must be 6 columns for the ordinal methods (IRV, Kemeny-Young, positional).
n	Size of electorate
window	Window within which two candidates are considered to be tied when method="rectangular". Wider window means lower variance but more bias.
cand_names	Names of the candidates.
sep	Separation between candidate names.
merge	Merge adjacent pivot events?
drop	Drop a dimension?
skip_non_pivot_events	Skip non pivot events?
s	Score allocated to a second-place ranking in positional and IRV elections.
raw	Return counts rather than integrals/proportions?
kemeny	For Condorcet method, compute Kemeny-Young pivot event probabilities? If F, only handles event in which the Condorcet winner is decided.

Details

For plurality elections, any number of candidates can be specified. The other voting systems can only handle three candidates.

Results have been validated against election_event_probs().

Examples

sims <- gtools::rdirichlet(100000, alpha = c(9,7,3,4,4,6))
plurality_pivot_probs_from_sims(sims)
positional_pivot_probs_from_sims(sims, s=.5)
irv_pivot_probs_from_sims(sims)
condorcet_pivot_probs_from_sims(sims)

aeggers/pivotprobs documentation built on Oct. 28, 2024, 9:46 a.m.