FA-12681 / Causal analysis data contracts / Open access

Crossover carryover coding forgets actual prior exposure across untreated periods · case 01

A carryover covariate follows assigned treatment or disappears during an untreated period without a documented washout.

Verified by executionVariant 1 · 8 checks per implementationDownload source bundle ↓JSON ↗

ROOT CAUSE

Sequence assignment or immediate previous receipt substitutes for persistent actual exposure history in the supplied crossover design.

VERIFIED REPAIR

Within each subject, process periods in order; clear history at documented washout, emit the pre-period carryover label, then update history only for actual treatment receipt.

Unsuccessful approach: Using immediate previous receipt corrects noncompliance but treats an untreated period as if it washed out prior exposure.

Case contract

Rows [subject,period,assigned,received,washed_out] have unique subject-period pairs. Treatments are A/B or None. The synthetic protocol defines carryover as the most recent actually received treatment since the last documented washout, persisting through untreated periods. Return sorted [subject,period,carryover] rows, with carryover evaluated before current receipt and after current washout. This is a categorical design encoding, not a pharmacokinetic claim.

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):
    history = {}
    result = []
    for subject, period, assigned, received, washed_out in sorted(rows):
        if washed_out:
            history[subject] = None
        result.append([subject, period, history.get(subject)])
        if assigned is not None:
            history[subject] = assigned
    return result
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('actual receipt overrides assignment', solve([['s',N,'A','B',False],['s',N+1,'A','A',False]]), [['s',N,None],['s',N+1,'B']])
check('untreated period preserves carryover', solve([['s',N,'A','A',False],['s',N+1,None,None,False],['s',N+2,'B','B',False]]), [['s',N,None],['s',N+1,'A'],['s',N+2,'A']])
check('washout clears before current receipt', solve([['s',N,'A','A',False],['s',N+1,'B','B',True],['s',N+2,None,None,False]]), [['s',N,None],['s',N+1,None],['s',N+2,'B']])
check('subjects have independent histories', solve([['s',N,'A','A',False],['t',N,'B','B',False],['s',N+1,None,None,False],['t',N+1,None,None,False]]), [['s',N,None],['s',N+1,'A'],['t',N,None],['t',N+1,'B']])
check('input order does not determine history', solve([['s',N+1,'B','B',False],['s',N,'A','A',False]]), [['s',N,None],['s',N+1,'A']])
check('no prior receipt means no carryover', solve([['s',N,'A',None,False],['s',N+1,'B','B',False]]), [['s',N,None],['s',N+1,None]])
check('empty design', solve([]), [])
check('washout without new receipt remains clear', solve([['s',N,'A','A',False],['s',N+1,None,None,True],['s',N+2,'B','B',False]]), [['s',N,None],['s',N+1,None],['s',N+2,None]])
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
actual receipt overrides assignment	`[['s', 1, None], ['s', 2, 'A']]`	`[['s', 1, None], ['s', 2, 'B']]`	Failed
untreated period preserves carryover	`[['s', 1, None], ['s', 2, 'A'], ['s', 3, 'A']]`	`[['s', 1, None], ['s', 2, 'A'], ['s', 3, 'A']]`	Passed
washout clears before current receipt	`[['s', 1, None], ['s', 2, None], ['s', 3, 'B']]`	`[['s', 1, None], ['s', 2, None], ['s', 3, 'B']]`	Passed
subjects have independent histories	`[['s', 1, None], ['s', 2, 'A'], ['t', 1, None], ['t', 2, 'B']]`	`[['s', 1, None], ['s', 2, 'A'], ['t', 1, None], ['t', 2, 'B']]`	Passed
input order does not determine history	`[['s', 1, None], ['s', 2, 'A']]`	`[['s', 1, None], ['s', 2, 'A']]`	Passed
no prior receipt means no carryover	`[['s', 1, None], ['s', 2, 'A']]`	`[['s', 1, None], ['s', 2, None]]`	Failed
empty design	`[]`	`[]`	Passed
washout without new receipt remains clear	`[['s', 1, None], ['s', 2, None], ['s', 3, None]]`	`[['s', 1, None], ['s', 2, None], ['s', 3, None]]`	Passed

SHA-256 / 9e317c78c0cc3ee0617d2fb2be24d81162d06efc01703c85fb3b95eb8255b607

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(rows):
    history = {}
    result = []
    for subject, period, assigned, received, washed_out in sorted(rows):
        if washed_out:
            history[subject] = None
        result.append([subject, period, history.get(subject)])
        history[subject] = received
    return result
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('actual receipt overrides assignment', solve([['s',N,'A','B',False],['s',N+1,'A','A',False]]), [['s',N,None],['s',N+1,'B']])
check('untreated period preserves carryover', solve([['s',N,'A','A',False],['s',N+1,None,None,False],['s',N+2,'B','B',False]]), [['s',N,None],['s',N+1,'A'],['s',N+2,'A']])
check('washout clears before current receipt', solve([['s',N,'A','A',False],['s',N+1,'B','B',True],['s',N+2,None,None,False]]), [['s',N,None],['s',N+1,None],['s',N+2,'B']])
check('subjects have independent histories', solve([['s',N,'A','A',False],['t',N,'B','B',False],['s',N+1,None,None,False],['t',N+1,None,None,False]]), [['s',N,None],['s',N+1,'A'],['t',N,None],['t',N+1,'B']])
check('input order does not determine history', solve([['s',N+1,'B','B',False],['s',N,'A','A',False]]), [['s',N,None],['s',N+1,'A']])
check('no prior receipt means no carryover', solve([['s',N,'A',None,False],['s',N+1,'B','B',False]]), [['s',N,None],['s',N+1,None]])
check('empty design', solve([]), [])
check('washout without new receipt remains clear', solve([['s',N,'A','A',False],['s',N+1,None,None,True],['s',N+2,'B','B',False]]), [['s',N,None],['s',N+1,None],['s',N+2,None]])
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
actual receipt overrides assignment	`[['s', 1, None], ['s', 2, 'B']]`	`[['s', 1, None], ['s', 2, 'B']]`	Passed
untreated period preserves carryover	`[['s', 1, None], ['s', 2, 'A'], ['s', 3, None]]`	`[['s', 1, None], ['s', 2, 'A'], ['s', 3, 'A']]`	Failed
washout clears before current receipt	`[['s', 1, None], ['s', 2, None], ['s', 3, 'B']]`	`[['s', 1, None], ['s', 2, None], ['s', 3, 'B']]`	Passed
subjects have independent histories	`[['s', 1, None], ['s', 2, 'A'], ['t', 1, None], ['t', 2, 'B']]`	`[['s', 1, None], ['s', 2, 'A'], ['t', 1, None], ['t', 2, 'B']]`	Passed
input order does not determine history	`[['s', 1, None], ['s', 2, 'A']]`	`[['s', 1, None], ['s', 2, 'A']]`	Passed
no prior receipt means no carryover	`[['s', 1, None], ['s', 2, None]]`	`[['s', 1, None], ['s', 2, None]]`	Passed
empty design	`[]`	`[]`	Passed
washout without new receipt remains clear	`[['s', 1, None], ['s', 2, None], ['s', 3, None]]`	`[['s', 1, None], ['s', 2, None], ['s', 3, None]]`	Passed

SHA-256 / e15ddac916e11ff7d0d9f684d0da90a1853d85772357c865f1fe7e7c9961a1b1

3 / The verified repair

Exit 0

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

N = 1
observations = []
def solve(rows):
    history = {}
    result = []
    for subject, period, assigned, received, washed_out in sorted(rows):
        if washed_out:
            history[subject] = None
        result.append([subject, period, history.get(subject)])
        if received is not None:
            history[subject] = received
    return result
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('actual receipt overrides assignment', solve([['s',N,'A','B',False],['s',N+1,'A','A',False]]), [['s',N,None],['s',N+1,'B']])
check('untreated period preserves carryover', solve([['s',N,'A','A',False],['s',N+1,None,None,False],['s',N+2,'B','B',False]]), [['s',N,None],['s',N+1,'A'],['s',N+2,'A']])
check('washout clears before current receipt', solve([['s',N,'A','A',False],['s',N+1,'B','B',True],['s',N+2,None,None,False]]), [['s',N,None],['s',N+1,None],['s',N+2,'B']])
check('subjects have independent histories', solve([['s',N,'A','A',False],['t',N,'B','B',False],['s',N+1,None,None,False],['t',N+1,None,None,False]]), [['s',N,None],['s',N+1,'A'],['t',N,None],['t',N+1,'B']])
check('input order does not determine history', solve([['s',N+1,'B','B',False],['s',N,'A','A',False]]), [['s',N,None],['s',N+1,'A']])
check('no prior receipt means no carryover', solve([['s',N,'A',None,False],['s',N+1,'B','B',False]]), [['s',N,None],['s',N+1,None]])
check('empty design', solve([]), [])
check('washout without new receipt remains clear', solve([['s',N,'A','A',False],['s',N+1,None,None,True],['s',N+2,'B','B',False]]), [['s',N,None],['s',N+1,None],['s',N+2,None]])
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
actual receipt overrides assignment	`[['s', 1, None], ['s', 2, 'B']]`	`[['s', 1, None], ['s', 2, 'B']]`	Passed
untreated period preserves carryover	`[['s', 1, None], ['s', 2, 'A'], ['s', 3, 'A']]`	`[['s', 1, None], ['s', 2, 'A'], ['s', 3, 'A']]`	Passed
washout clears before current receipt	`[['s', 1, None], ['s', 2, None], ['s', 3, 'B']]`	`[['s', 1, None], ['s', 2, None], ['s', 3, 'B']]`	Passed
subjects have independent histories	`[['s', 1, None], ['s', 2, 'A'], ['t', 1, None], ['t', 2, 'B']]`	`[['s', 1, None], ['s', 2, 'A'], ['t', 1, None], ['t', 2, 'B']]`	Passed
input order does not determine history	`[['s', 1, None], ['s', 2, 'A']]`	`[['s', 1, None], ['s', 2, 'A']]`	Passed
no prior receipt means no carryover	`[['s', 1, None], ['s', 2, None]]`	`[['s', 1, None], ['s', 2, None]]`	Passed
empty design	`[]`	`[]`	Passed
washout without new receipt remains clear	`[['s', 1, None], ['s', 2, None], ['s', 3, None]]`	`[['s', 1, None], ['s', 2, None], ['s', 3, None]]`	Passed

SHA-256 / e59219fe9526b33b933bb63b7ecb77db44d1cb65b546e06a9f73815e50b5adf2

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 2026-09-29T14:38:59.175403+00:00.

Case digest / 915c91c8f06f057cf89c9aeb4b700ef2b5bfc4ca30af701b437864e0d6fdf5b0