Log-likelihood matrix for the simple Bradley–Terry model

Builds an T_iter x D matrix of log-likelihood values, where T_iter is the number of posterior draws and D is the number of observed unordered pairs (i<j) with n_ij > 0. This is suitable as input to loo.

Builds an \(S \times D\) matrix of log-likelihood values, where \(S\) is the number of posterior draws and \(D\) is the number of observed unordered pairs (i<j) with \(n_{ij} > 0\).

Usage

make_bt_simple_loo(w_ij, lambda_samples)

make_bt_simple_loo(w_ij, lambda_samples)

Arguments

w_ij

integer/numeric \(n \times n\) wins (i over j).

lambda_samples

numeric \(S \times n\) matrix of player-specific rates \(\lambda_i\).

Value

A list with:

• ll — T_iter x D matrix of log-likelihoods.

• obs_idx — D x 2 matrix of (i,j) indices defining each column.

A list with:

• ll — \(S \times D\) matrix of log-likelihoods.

• obs_idx — \(D \times 2\) matrix of (i,j) indices defining each column.