Struct DesignedRowSample 

pub struct DesignedRowSample {
    pub provenance: MeasureProvenance,
    pub rows: Vec<usize>,
    pub likelihood_weights: Vec<f64>,
    pub expected_size: f64,
}

A designed importance subsample with honest Horvitz–Thompson likelihood weights — what a frontier fit sums over instead of the full corpus (#987 / #973). Produced by RowSamplingMeasure::designed_subsample.

Fields

provenance: MeasureProvenance

Provenance of the measure that shaped the design (uniform fallback or Fisher mass), echoed for consumer certification.

rows: Vec<usize>

Selected row indices, ascending.

likelihood_weights: Vec<f64>

Per-selected-row likelihood weight 1 / π_i, aligned with rows. Multiplying row i’s loss term by this makes the subsampled criterion unbiased for the full-corpus criterion.

expected_size: f64

Σ π_i — the design’s expected sample size (≈ the requested budget; Madow selection realizes ⌊·⌋ or ⌈·⌉ of it).

Implementations

impl DesignedRowSample

pub fn len(&self) -> usize

Number of rows actually selected.

pub fn is_empty(&self) -> bool

pub fn estimated_corpus_rows(&self) -> f64

Σ 1/π_i over the selected rows — the Horvitz–Thompson estimate of the corpus row count. A consumer can sanity-gate the design by checking this lands near n (it is exactly n in expectation).

Trait Implementations

impl Clone for DesignedRowSample

fn clone(&self) -> DesignedRowSample

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for DesignedRowSample

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

Formats the value using the given formatter.