Simple Bradley–Terry Gibbs sampler (no clustering)

Simple Bradley–Terry Gibbs sampler (no clustering)

Usage

gibbs_bt_simple(
  w_ij,
  a = 0.01,
  b = 0.1,
  T_iter = 5000,
  T_burn = 1000,
  verbose = TRUE
)

Arguments

w_ij

integer/numeric \(n \times n\) wins from i over j (diag = 0, nonnegative).

a, b

numeric(1) Gamma(a,b) prior on each \(\lambda_i\).

T_iter, T_burn

integers; total iterations and burn-in. Require T_burn < T_iter.

verbose

logical; print progress every 1000 iterations.

Value

A list with lambda_samples (matrix of size \(S \times n\), \(S=T_{\text{iter}}-T_{\text{burn}}\)).

Examples

if (FALSE) { # \dontrun{
set.seed(1)
n <- 6L
w <- matrix(0L, n, n)
w[upper.tri(w)] <- rpois(sum(upper.tri(w)), 2)
w <- w + t(w) - diag(diag(w))
fit <- gibbs_bt_simple(w, a = 1, b = 1, T_iter = 500, T_burn = 100, verbose = FALSE)
} # }