Struct RankedRow 

pub struct RankedRow {
    pub name: String,
    pub score: f64,
    pub delta: f64,
    pub evidence_ratio: f64,
    pub edf: f64,
}

Fields

name: String

score: f64

delta: f64

Cost gap from the winning model on the SAME scale used to order the ranking (ranking_score, the Occam-penalised conditional AIC, issue #1362). The winner is argmin ranking_score, so this is >= 0 for every row by construction — it never contradicts the declared winner (issue #1465). score still carries the raw REML/LAML diagnostic on its own explicitly labelled scale.

evidence_ratio: f64

Akaike evidence ratio of the winner over this row on the ranking scale. delta is a conditional-AIC gap (a −2·log / deviance-scale quantity), so the evidence ratio is exp(½·delta) >= 1 (Burnham & Anderson), NOT exp(delta) — the latter squares the intended ratio (issues #1465, #2124).

This is NOT a Bayes factor and was renamed away from that word: a Bayes factor is a ratio of prior-integrated marginal likelihoods, whereas this is the relative likelihood exp(−ΔAIC/2), which integrates over no prior and must not be read against Jeffreys / Kass–Raftery thresholds. The raw REML/LAML score_table keeps the Laplace-approximate marginal-likelihood diagnostic on its own labelled scale.

edf: f64

Trait Implementations

impl Clone for RankedRow

fn clone(&self) -> RankedRow

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for RankedRow

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.