Agreement index A based on entropy of weights

For the retained set \(\mathcal B = \{i : w_i > 0\}\), computes entropy \(H(\tilde{\mathbf w}) = -\sum_{i\in \mathcal B}\tilde w_i \log \tilde w_i\) and \(A = H(\tilde{\mathbf w}) / \log|\mathcal B|\), with the convention \(A=0\) when \(|\mathcal B| \le 1\).

Usage

calc_model_agreement_index(w)

Arguments

w

Numeric vector of weights (raw or normalized). Only strictly positive entries form \(\mathcal B\). Internally normalized within \(\mathcal B\); larger \(A\) is better.

Value

A list with A (agreement index in [0,1]), H (entropy), and B_size (retained set size).

See also

calc_model_cvw()

Other calibration-metrics: calc_convergence_diagnostics(), calc_model_cvw(), calc_model_ess()