Reject a subject-wide count inside a one-race row

Review ask

As a clinical trial statistician or programmer, answer two short questions in Your review: the task and the inputs. Agreeing with both approves this benchmark; disagreeing with either needs a comment. Propose a fix with the Edit button on any source file, or add it to your comments.

We suggest spending no more than 5 minutes on a review. The other tabs beside Your review are background for assessment, not under review.

Summary

EditUnder reviewWhat this benchmark means

README.md

Public entries: Python and R. Passing coverage requires an installed inventory.

Goal: attempt one supplemental Demographics (SUPPDM) record per reported race (QVAL) for each subject who marked more than one race, deciding "more than one" by counting the subject's race rows.

Input: EDC output in long form, one row per collected item; each race option a subject marks contributes one race row.

Variables:

QVAL: would be one race the subject marked, kept only for a subject with several.

Each record being built stands for one subject and one race, so a count taken while building it sees only that race's rows: a subject with two different races would count one for each and be dropped, while a race marked twice would count two and be kept. Asking the count to span the whole subject does not widen it, so the run is rejected before any data is read.

How to fix

Decide which record already says that a subject marked several races, and read it instead of recounting inside the record being built. When the same run builds DM, its RACE is MULTIPLE exactly then: read DM through an input with schema: and filter on DMRACE = 'MULTIPLE', as sdtm-dm-race-ethnicity does.

Without DM, select the subject's lowest and highest race with keyed lookups and keep the record when they differ. Unlike a row count, this does not mistake a race marked twice for two races:

intermediates:
  - id: RACELOW
    dataset: ODM
    key: {StudyOID: STUDYID, SubjectKey: USUBJID}
    filter: "ODM.ItemOID = 'IT.DM.RACE' AND ODM.Value IS NOT NULL"
    order_by: [ODM.Value]
    keep: first
  # RACEHIGH: the same lookup with keep: last
rows:
  - id: reported_race
    group_by: [ODM.StudyOID, ODM.SubjectKey, ODM.ItemOID, ODM.Value]
    filter: "ITEMOID = 'IT.DM.RACE' AND RACELOWV < RACEHIGHV"
    derivations:
      # STUDYID, USUBJID, ITEMOID and QVAL as before
      RACELOWV: RACELOW.Value
      RACEHIGHV: RACEHIGH.Value

Declare RACELOWV and RACEHIGHV as text columns in place of RACECOUNT, and leave them out of the output.

Expected failure

Editthe run is rejected; no artifact is accepted

expected/error.yaml

phase: validation
condition: invalid_aggregate_context
rule: LOOKUP-5

expected/error.yaml

phase: validation
condition: invalid_aggregate_context
spec_paths:
  - rows.reported_race.derivations.RACECOUNT.aggregate.key
rule: LOOKUP-5
context:
  expr: "COUNT(ODM.*)"
  reason: a grouped row aggregate reads its own group and declares no key pairs

YAML specification

1 spec file

spec.yaml

Edit52 lines

1yamaa: "1.0"2type: spec3model: SDTM4domain: SUPPDM5keys: [STUDYID, USUBJID, QVAL]67input:8  ODM: input/odm.csv910output:11  path: suppdm.csv12  columns: [STUDYID, USUBJID, QVAL]1314columns:15  - name: STUDYID16    type: str17    label: Study Identifier1819  - name: USUBJID20    type: str21    label: Unique Subject Identifier2223  - name: ITEMOID24    type: str25    label: Collected Item2627  - name: QVAL28    type: str29    label: Data Value3031  - name: RACECOUNT32    type: int33    label: Number of Race Rows for the Subject3435rows:36  # Each candidate is one subject and one collected value, so the count below37  # can only ever see this group's own records. The key pair asks for the38  # subject's race rows instead; a grouped row reduction cannot widen its39  # read, so the specification is rejected rather than counting one.40  - id: reported_race41    group_by: [ODM.StudyOID, ODM.SubjectKey, ODM.ItemOID, ODM.Value]42    filter: "ITEMOID = 'IT.DM.RACE' AND RACECOUNT > 1"43    derivations:44      STUDYID: ODM.StudyOID45      USUBJID: ODM.SubjectKey46      ITEMOID: ODM.ItemOID47      QVAL: ODM.Value48      RACECOUNT:49        aggregate:50          filter: "ODM.ItemOID = 'IT.DM.RACE'"51          expr: "COUNT(ODM.*)"52          key: {StudyOID: STUDYID, SubjectKey: USUBJID}

Inputs

Under review1 source file

input/odm.csv

Edit5 rows

input/odm.csv
StudyOID	MetaDataVersionOID	SubjectKey	StudyEventOID	StudyEventRepeatKey	ItemGroupOID	ItemGroupRepeatKey	ItemOID	Value
STUDY01	MDV.1	001	SCREENING	1	IG.DM	1	IT.DM.RACE	White
STUDY01	MDV.1	002	SCREENING	1	IG.DM	1	IT.DM.RACE	Asian
STUDY01	MDV.1	002	SCREENING	1	IG.DM	1	IT.DM.RACE	Black or African American
STUDY01	MDV.1	003	SCREENING	1	IG.DM	1	IT.DM.RACE	White
STUDY01	MDV.1	003	SCREENING	1	IG.DM	1	IT.DM.RACE	White

Unexpected Output

Under reviewNo artifact is produced

No fixture files in this directory.

Shaded columns are derived— = missing value

Human review and assessment

Variable Name	Variable Label	Type	Origin	Conversion Definition	Variable Type	Variable Order
STUDYID	Study Identifier	Char	Derived	Set per constructed row; see the Row construction sheet.	SUPP	1
USUBJID	Unique Subject Identifier	Char	Derived	Set per constructed row; see the Row construction sheet.	SUPP	2
QVAL	Data Value	Char	Derived	Set per constructed row; see the Row construction sheet.	SUPP	3

Template	Include when	STUDYID	USUBJID	ITEMOID	QVAL	RACECOUNT
reported_race	ITEMOID = 'IT.DM.RACE' AND RACECOUNT > 1	Copy ODM.StudyOID	Copy ODM.SubjectKey	Copy ODM.ItemOID	Copy ODM.Value	Aggregate COUNT(ODM.*) over records whose StudyOID equals STUDYID and SubjectKey equals USUBJID, where ODM.ItemOID = 'IT.DM.RACE'

Version	Date	Author	Description
1.7	2026-10-20	yamaa docs build	Read a specification in the 1.0 surface as its 0.5 counterpart; state that missing values stay missing where a mapping declares no missing handler, and leave out a declared missing answer that restates a handler's default
1.6	2026-10-17	yamaa docs build	Describe collected date parts written as ISO 8601 date text and times of day counted in seconds since midnight
1.5	2026-10-17	yamaa docs build	Name collected precision in words instead of letter codes; state imputation fallbacks, minimum source precision, bounds ignored outside the collected period, and invalid text with no handler as an error; describe date and datetime shifts, imputation flag codes, and partial text giving no date part
1.4	2026-10-15	yamaa docs build	State datetime precision codes through the hour and minute, and complete only the time parts a collected datetime lacks
1.3	2026-09-29	yamaa docs build	Show study codelist names, external dictionaries, and ISO 8601 formats in variable rows; list the document's codelist values and metadata
1.2	2026-09-29	yamaa docs build	Classify ADaM copies from a declared input as Predecessor and recodes as Derived, and SDTM recodes of locally derived columns as Derived
1.1	2026-09-29	yamaa docs build	Resolve lookups into the donor record they select, and give every derivation shape a plain-language rule instead of its raw YAML
1.0		yamaa docs build	Initial generation from spec.yaml

How to fix

input/odm.csv

Your review

Mapping spec