Quantify incremental process information

Compares posterior draws for the same latent target under a baseline and an augmented model. This implementation intentionally uses posterior uncertainty metrics rather than assuming Fisher information is additively decomposable across heterogeneous channels.

Usage

process_information(
  baseline,
  augmented,
  metric = c("variance_reduction", "precision_gain", "entropy_reduction")
)

Arguments

baseline, augmented

Numeric posterior draws. Rows are draws and columns are matched latent targets.

metric

`variance_reduction`, `precision_gain`, or `entropy_reduction`.

Value

An `eye_process_information` data.frame.