FA-12661 / Causal analysis data contracts / Open access

Cluster randomization accepts contradictory unit assignments · case 01

Cluster randomization accepts contradictory unit assignments.

Verified by executionVariant 1 · 6 checks per implementationDownload source bundle ↓JSON ↗

ROOT CAUSE

A cluster is treated as a set of independent unit assignments.

THE FAILURE

A cluster is treated as a set of independent unit assignments.

Unsuccessful approach: Majority vote masks minority contradictions.

Case contract

Rows [cluster,arm] use arms 0 and 1. Return sorted clusters violating the cluster-level assignment design.

Why this case matters

A deterministic synthetic study model isolates this data-contract defect; outputs alone establish no real-world causal identification.

1 / The failure

Exit 1

"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(rows):
    return []
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('unequal conflict', solve([['a',0]]*(N+1)+[['a',1]]), ['a'])
check('balanced conflict', solve([['a',0],['a',1]]), ['a'])
check('all treated', solve([['a',1]]*N), [])
check('all control', solve([['a',0]]*N), [])
check('two clusters', solve([['a',0],['b',1]]), [])
check('empty', solve([]), [])
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)

Boundary fixture	Actual	Expected	Outcome
unequal conflict	`[]`	`['a']`	Failed
balanced conflict	`[]`	`['a']`	Failed
all treated	`[]`	`[]`	Passed
all control	`[]`	`[]`	Passed
two clusters	`[]`	`[]`	Passed
empty	`[]`	`[]`	Passed

SHA-256 / 06144e31a0f7998b711e71cd52e76b99205e3e5fd45df9c82ab5dffd68cfe89f

2 / The unsuccessful fix

Exit 1

"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(rows):
    clusters = {c for c,a in rows}
    return sorted(c for c in clusters if sum(a for k,a in rows if k == c)*2 == sum(k == c for k,a in rows))
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('unequal conflict', solve([['a',0]]*(N+1)+[['a',1]]), ['a'])
check('balanced conflict', solve([['a',0],['a',1]]), ['a'])
check('all treated', solve([['a',1]]*N), [])
check('all control', solve([['a',0]]*N), [])
check('two clusters', solve([['a',0],['b',1]]), [])
check('empty', solve([]), [])
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)

Boundary fixture	Actual	Expected	Outcome
unequal conflict	`[]`	`['a']`	Failed
balanced conflict	`['a']`	`['a']`	Passed
all treated	`[]`	`[]`	Passed
all control	`[]`	`[]`	Passed
two clusters	`[]`	`[]`	Passed
empty	`[]`	`[]`	Passed

SHA-256 / 9b804499f46999c881cc63da501c084a3afc8f2c35ef8d5117f945595d362e9e

HELD IN THE MEMBER ARCHIVE

The verified repair and its recorded checks are member-only.

This mechanism has 6 recorded checks per implementation. The open-access tier publishes the failure and the unsuccessful fix; the repaired source that passes every check, and the observations that prove it, are available to members.

Every case sharing this mechanism uses the same contract and the same repair, so this one record is held back for all of them.

Member access is invitation-based. Sign in with your invited account to inspect the repair.

Verification & scope

Finite, fully specified synthetic data only; identification assumptions are supplied by the fixture design, not inferred from observations. This reproducer isolates one failure mechanism. Results cover the supplied fixtures. Variants within a family share a test contract and should remain grouped when constructing evaluation splits. Related mechanisms with a shared evaluation_group must also remain together; these controlled models are not independent production incidents.

Observations recorded using Python 3.12.14 at 2026-09-29T14:38:58.834900+00:00.

Case digest / ce4838fc3bfe931ec71221f5377ee4293480a50df8233cf1478e28333b0bd4bd