plackett_luce: Gibbs Sampler for posterior plackett luce parameters

In dcbdan/s525: Implementation of common Bayesian models

Description

The prior on each parameter is given by a gamma distribution.

The parameter expansion is given by Caron, Francois, and Arnaud Doucet. "Efficient Bayesian inference for generalized Bradley<c2><bf>Terry models." Journal of Computational and Graphical Statistics 21.1 (2012): 174-196.

Usage

plackett_luce <- function(
  rank_matrix,
  shape = 1,
  rate = 1,
  niter = 1000)

Arguments

rank_matrix	n by p matrix of integers ranging from 1,...,k where k = max(rank_matrix, na.rm=TRUE). Each row may contain NA values provided that at least 2 values of each row are not NA. The same integer can appear in each row of the rank matrix as well.
shape	scalar or k vector specifying prior shape parameter over pl parameters
rate	scalar or k vector specifying prior rate parameter over pl parameters
niter	number of gibbs sampling iterations

Details

This function returns a niter x k matrix of the gibbs sampler.

Examples

  library(coda)

  vals <- rplackett_luce(100, vs = c(100, 1, 1))
  post_vals <- plackett_luce(vals)
  plot(as.mcmc(post_vals))

dcbdan/s525 documentation built on May 19, 2019, 10:48 p.m.