Stratified Partition Sampling Check¶
Downstream fitness audit — instantiates Exhaustive Disjoint Partition Design
Certifies that a partition is safe to use as sampling strata — every unit in exactly one stratum and the strata covering the whole frame — before any estimate is drawn from it.
Stratified Partition Sampling Check is the acceptance gate a downstream consumer runs before treating a partition as sampling strata. Stratified estimation is only unbiased when the strata are mutually exclusive and collectively exhaustive of the sampling frame: a unit in two strata gets double weight, a unit in none never gets drawn, and either way the estimate is quietly wrong. This check verifies both invariants for the specific requirements sampling imposes and reconciles the stratum sizes against the known frame total. Its distinctive stance is that it validates the partition at the point of use, on behalf of the estimate about to be computed — it is the consumer-side enforcement of the fitness contract, not the producer-side maintenance of the scheme.
Example¶
An official household survey is stratified by region crossed with urbanicity. Before a single household is drawn, the sampling check runs against the frame: does every household map to exactly one stratum, and do the strata's household counts sum to the full frame? It catches a boundary town coded into two adjacent strata — those households would be over-represented — and a recently annexed district that fell into no stratum at all, a coverage error that would bias every regional estimate downward. Because the check is a gate, the sample isn't drawn until the frame is fixed, so the design weights that come out are valid rather than subtly broken.
How it works¶
- Run the two invariants as a gate. Confirm each frame unit sits in exactly one stratum (disjointness) and that the union of strata equals the frame (exhaustiveness) before sampling, not after.
- Reconcile weights to the frame. Check that stratum sizes or measures-of-size sum to the known frame total; a mismatch is a gap or an overlap surfacing as arithmetic.
- Report fitness, not a fix. Emit a pass/fail for this use — weights valid, safe to sample — and send any failure back to the partition's stewards rather than patching it in place.
Tuning parameters¶
- Frame definition — what population counts as the universe the strata must cover. A stale or wrong frame can pass the check while the sample quietly targets the wrong world.
- Tolerance — exact reconciliation vs an allowed small unclassified share. Loosening lets fieldwork proceed but injects bias into the estimates.
- Weight basis — count-based vs measure-of-size weights, which changes what "exhaustive" must sum to and how a shortfall reads.
- Gate strictness — a hard stop on any violation vs warn-and-proceed with an adjustment factor. Hard stops protect validity; soft gates protect the schedule.
- Re-check trigger — once per sample design vs every wave. Per-wave catches frame drift over time but costs recurring effort.
When it helps, and when it misleads¶
Its strength is that it makes MECE a precondition of inference rather than an afterthought: it refuses to draw a sample on a partition that would bias the result, and it localizes the fault (overlap → over-weighting; gap → coverage error) so stewards know what to fix. The unbiasedness of stratified sampling rests directly on the strata being mutually exclusive and exhaustive of the frame, and this check is simply that precondition enforced at the door.[n1]
Its blind spot is that it validates the strata against the frame, so a wrong or outdated frame sails through while the sample misses reality — coverage error that lives in the frame, not in the partition, is invisible here. The classic misuse is relaxing the tolerance until a convenient partition "passes," then reporting the estimates as if they were unbiased. The discipline is to tie the check to a current, defensible frame, keep the gate hard, and treat a failure as a reason to repair the partition — never as a reason to loosen the check.
How it implements the components¶
pairwise_disjointness_invariant— verifies each frame unit falls in exactly one stratum, so no unit is double-weighted.collective_exhaustiveness_invariant— verifies the strata cover the entire frame, so no unit is structurally un-sampleable.downstream_block_use_contract— it is the acceptance gate where the sampling consumer enforces the fitness the partition must meet before being relied upon.
It does not build the strata or their membership rules (block_membership_criteria, partition_construction_basis) — that's Block Membership Decision Table and Equivalence-Class Partition Derivation — and it does not translate strata across versions (partition_version_crosswalk); that's Partition Crosswalk Table.
Related¶
- Instantiates: Exhaustive Disjoint Partition Design — it is where the design's MECE guarantee is cashed in for valid estimation.
- Consumes: Overlap Detection Join and Coverage Gap Audit — it applies their disjointness and coverage tests, scoped to the sampling frame and run as a gate.
- Sibling mechanisms: Overlap Detection Join · Coverage Gap Audit · Block Membership Decision Table · Equivalence-Class Partition Derivation · MECE Partition Template
Editorial Notes¶
Form Classification¶
Form family: Assessment, Review & Assurance
Rationale: Stratified Partition Sampling Check operates as a bounded evaluation of existing evidence or work that produces a finding or disposition because it certifies that a partition is safe to use as sampling strata — every unit in exactly one stratum and the strata covering the whole frame — before any estimate is drawn from it.
Independent corroboration: The frozen evidence defines Stratified Partition Sampling Check as 'Certifies that a partition is safe to use as sampling strata — every unit in exactly one stratum and the strata covering the whole frame — before any estimate is drawn from it', so its operative form is Assessment, Review & Assurance.
Nearest alternative: Rule, Policy & Commitment — Stratified Partition Sampling Check includes features of a standing rule, threshold, contractual commitment, or policy constraint governing future conduct, but its defining operation is a bounded evaluation of existing evidence or work that produces a finding or disposition.
Review outcome: Independent reviewer agreement; medium confidence.
Origin Attribution¶
Primary origin: Statistics & Experimental Design
Origin pattern: Single lineage
Present-day reach: Specialized
Rationale: Verifying exhaustive disjoint strata is sampling-frame validation.
Related originating lineages:
- Data Science & Analytics — Data science, analytics, and operational monitoring supplies a parallel or contributing lineage for the mechanism's defining operation: certifies that a partition is safe to use as sampling strata — every unit in exactly one stratum and the strata covering the whole frame — before any estimate is drawn from it.
- Mathematics — A partition is disjoint and covering.
Review resolution: The blind reviewers agree that statistics_experimental_design is the primary origin and differ only on alternate origin disagreement. I preserve every independently explained alternate from both records rather than imposing a numeric cap. I retain single_lineage because the combined evidence shows one traceable formative lineage. The broader reach of specialized records portability separately from historical provenance; encyclopedia_synthesis=false preserves the affirmative synthesis judgment where either reviewer identified one.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
[n1] In stratified sampling the population frame is divided into strata that must be mutually exclusive and collectively exhaustive; unbiased estimation and correct design weights depend on that MECE structure holding. Overlap inflates some units' representation, and a gap silently excludes part of the population — which is why the check runs before, not after, the draw. ↩