Struct Posterior 

pub struct Posterior {

    pub delta_post: Vector9,
    pub lambda_post: Matrix9,
    pub beta_proj: Vector2,
    pub beta_proj_cov: Matrix2,
    pub beta_draws: Vec<Vector2>,
    pub leak_probability: f64,
    pub projection_mismatch_q: f64,
    pub n: usize,
    pub lambda_mean: Option<f64>,
    pub lambda_mixing_ok: Option<bool>,
    pub kappa_mean: Option<f64>,
    pub kappa_cv: Option<f64>,
    pub kappa_ess: Option<f64>,
    pub kappa_mixing_ok: Option<bool>,
}

Posterior distribution parameters for the 9D effect vector δ.

The posterior is Gaussian: δ | Δ ~ N(δ_post, Λ_post) where each δ_k represents the timing difference at decile k.

The 2D projection β_proj = (μ, τ) provides shift/tail decomposition.

Uses Student’s t prior (ν=4) via Gibbs sampling for robust inference.

Fields

delta_post: Vector9

9D posterior mean δ_post in nanoseconds.

lambda_post: Matrix9

9D posterior covariance Λ_post.

beta_proj: Vector2

2D GLS projection β = (μ, τ) for interpretability.

• β[0] = μ (uniform shift)
• β[1] = τ (tail effect)

beta_proj_cov: Matrix2

2D projection covariance (empirical, from draws).

beta_draws: Vec<Vector2>

Retained β draws for dominance-based classification (spec §3.4.6). Each draw is β^(s) = A·δ^(s) where A is the GLS projection matrix.

leak_probability: f64

Leak probability: P(max_k |δ_k| > θ | Δ). Computed via Monte Carlo integration over the 9D posterior.

projection_mismatch_q: f64

Projection mismatch Q statistic. Q = r’Σ_n⁻¹r where r = δ_post - X β_proj. High values indicate the 2D summary is approximate.

n: usize

Number of samples used in this posterior computation.

lambda_mean: Option<f64>

Posterior mean of latent scale λ. None if using simple posterior (no Gibbs sampler).

lambda_mixing_ok: Option<bool>

Whether the Gibbs sampler’s lambda chain mixed well. None if using simple posterior. When Some(false), indicates potential posterior unreliability.

kappa_mean: Option<f64>

Posterior mean of likelihood precision κ. None if using simple posterior.

kappa_cv: Option<f64>

Coefficient of variation of κ. None if using simple posterior.

kappa_ess: Option<f64>

Effective sample size of κ chain. None if using simple posterior.

kappa_mixing_ok: Option<bool>

Whether the Gibbs sampler’s kappa chain mixed well. None if using simple posterior.

Implementations

impl Posterior

pub fn new( delta_post: Vector9, lambda_post: Matrix9, beta_proj: Vector2, beta_proj_cov: Matrix2, beta_draws: Vec<Vector2>, leak_probability: f64, projection_mismatch_q: f64, n: usize, ) -> Self

Create a new posterior with given parameters.

pub fn new_with_gibbs( delta_post: Vector9, lambda_post: Matrix9, beta_proj: Vector2, beta_proj_cov: Matrix2, beta_draws: Vec<Vector2>, leak_probability: f64, projection_mismatch_q: f64, n: usize, lambda_mean: f64, lambda_mixing_ok: bool, kappa_mean: f64, kappa_cv: f64, kappa_ess: f64, kappa_mixing_ok: bool, ) -> Self

Create a new posterior with Gibbs sampler diagnostics (v5.4, v5.6).

pub fn shift_ns(&self) -> f64

Get the shift component (μ) from the 2D projection.

pub fn tail_ns(&self) -> f64

Get the tail component (τ) from the 2D projection.

pub fn shift_se(&self) -> f64

Get the standard error of the shift component.

pub fn tail_se(&self) -> f64

Get the standard error of the tail component.

pub fn max_effect_ns(&self) -> f64

Get the max absolute effect across all deciles.

pub fn to_effect_estimate(&self) -> EffectEstimate

Build an EffectEstimate from this posterior.

This computes the credible interval and classifies the effect pattern using dominance-based classification from the β draws (spec §3.4.6).

pub fn measurement_quality(&self) -> MeasurementQuality

Get measurement quality based on the effect standard error.

Quality is determined by the minimum detectable effect (MDE), which is approximately 2x the effect standard error.

pub fn to_summary(&self) -> PosteriorSummary

Convert to an FFI-friendly summary containing only scalar fields.

This uses the canonical effect pattern classification from draws (spec §3.4.6) rather than the simpler heuristics that were previously duplicated in bindings.

Trait Implementations

impl Clone for Posterior

fn clone(&self) -> Posterior

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Posterior

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

Formats the value using the given formatter.