BayesMallows (version 0.3.0)

generate_initial_ranking: Generate Initial Ranking

Description

Given a consistent set of pairwise preferences, generate a complete ranking of items which is consistent with the preferences.

Usage

generate_initial_ranking(tc, n_items = max(tc[, c("bottom_item",
  "top_item")]), cl = NULL)

Arguments

tc

A dataframe with pairwise comparisons of S3 subclass BayesMallowsTC, returned from generate_transitive_closure.

n_items

The total number of items. If not provided, it is assumed to equal the largest item index found in tc, i.e., max(tc[, c("bottom_item", "top_item")]).

cl

Optional computing cluster used for parallelization, returned from parallel::makeCluster. Defaults to NULL.

Value

A matrix of rankings which can be given in the rankings argument to compute_mallows.

Examples

# NOT RUN {
# The example dataset beach_preferences contains pairwise prefences of beach.
# We must first generate the transitive closure
beach_tc <- generate_transitive_closure(beach_preferences)

# Next, we generate an initial ranking
beach_init <- generate_initial_ranking(beach_tc)

# Look at the first few rows:
head(beach_init)

# We can add more informative column names in order
# to get nicer posterior plots:
colnames(beach_init) <- paste("Beach", seq(from = 1, to = ncol(beach_init), by = 1))
head(beach_init)

# }
# NOT RUN {
  # We now give beach_init and beach_tc to compute_mallows. We tell compute_mallows
  # to save the augmented data, in order to study the convergence.
  model_fit <- compute_mallows(rankings = beach_init,
                               preferences = beach_tc,
                               nmc = 2000,
                               save_aug = TRUE)

  # We can study the acceptance rate of the augmented rankings
  assess_convergence(model_fit, parameter = "Rtilde")

  # We can also study the posterior distribution of the consensus rank of each beach
  model_fit$burnin <- 500
  plot(model_fit, parameter = "rho", items = 1:15)
# }
# NOT RUN {
# }
# NOT RUN {
  # The computations can also be done in parallel
  library(parallel)
  cl <- makeCluster(detectCores() - 1)
  beach_tc <- generate_transitive_closure(beach_preferences, cl = cl)
  beach_init <- generate_initial_ranking(beach_tc, cl = cl)
  stopCluster(cl)
# }