Create validated Plackett–Luce ranking data — as_rankings • BTSBM: Bayesian paired-comparison and ranking models

Rows of rankings are preference orderings and columns are successive positions, from most to least preferred. Each row may be a complete ranking or a top-m prefix from one common universe of items. Internally this is represented in the compact integer format used by the PLuce reference code, but users never need to work with that internal name.

Usage

as_rankings(
  rankings,
  item_count = NULL,
  ranking_labels = NULL,
  item_labels = NULL
)

as_pl_data(rho, n_items = NULL, ranking_id = NULL, item_labels = NULL)

Arguments

rankings: Integer-like matrix. Row l lists item ids from most to least preferred in ranking l.
item_count: Total number of available items. If NULL, it is inferred as max(rankings); supply it when a top-m dataset has never-ranked items.
ranking_labels: Optional unique character vector identifying ranking rows.
item_labels: Optional unique character vector of length item_count.
rho: Deprecated name for rankings.
n_items: Deprecated name for item_count.
ranking_id: Deprecated name for ranking_labels.

Value

An object of class btsbm_ranking_data.

References

Caron, F. and Doucet, A. (2012). Efficient Bayesian inference for generalized Bradley–Terry models. Journal of Computational and Graphical Statistics, 21(1), 174–196. doi:10.1080/10618600.2012.638220