Subject Dose Aggregation

Review ask

As a clinical trial statistician or programmer, challenge anything in this benchmark: describe a problem in Comments, or propose a fix with the Edit button on any source file.

Task: Would a real study need this derivation, specified this way? Does it follow CDISC ADaM: variable names, controlled terms, record structure? Flag any rule that is ambiguous or that a sponsor would handle differently.
Inputs: Do they look like data collected in a real trial, and do they include the situations you would expect to meet in one?
Expected output: Would independent QC programming reproduce every value from these inputs and rules?

The other tabs beside Comments are background for assessment, not under review.

Summary

EditUnder reviewWhat this benchmark means

README.md

Goal: summarize each subject's dosing records three ways: a qualified reduction over the dosing table, a filtered and a per-record-derived variant of it, and an unqualified reduction over the built rows that is broadcast back to every row in its treatment group.

Input: DM carries one row per subject with a treatment group (TRTGRP). EX carries the dosing records: subject, sequence, dose, and treatment. Subject 04 has no dosing records; subject 03 has one record whose dose is missing; subject 01 has three records, one with a missing dose.

Aggregations:

NDOSE counts dose records (COUNT(EX.*)); NDOSEVAL counts only the records with a dose present (COUNT(EX.EXDOSE)). TOTDOSE, MEANDOSE, and MAXDOSE are the total, mean and largest of the present doses.
DRUGDOSE sums only the DRUG records; PLACDOSE sums only the PLACEBO records, so subjects with no placebo record read as missing.
TOTDOSE_MG binds each record's dose in milligrams first (a derive step whose second binding reads the first) and then sums the bound values.
AVGNDOSE is unqualified: it averages NDOSE over the built rows within each treatment group and broadcasts the result to every row of the group. Subject 04's missing NDOSE is excluded from the average.

Note: a missing dose contributes nothing and never a zero: TOTDOSE for subject 03 is missing while NDOSEVAL is 0. A subject with no records at all (subject 04) reads every qualified reduction as missing, including both counts. An empty group after filtering (PLACDOSE for subject 01) also reads as missing.

YAML specification

1 spec file

spec.yaml

Edit117 lines

1schema_version: "1.0"2domain: ADSL3keys: [USUBJID]4input:5  DM: input/dm.csv6  EX: {path: input/ex.csv, types: {EXDOSE: float}}7base: DM89output:10  path: adsl.csv11  columns: [USUBJID, TRTGRP, NDOSE, NDOSEVAL, TOTDOSE, MEANDOSE, MAXDOSE,12            DRUGDOSE, PLACDOSE, TOTDOSE_MG, AVGNDOSE]1314columns:15  - name: USUBJID16    type: str17    label: Unique Subject Identifier18    derivation: DM.USUBJID1920  - name: TRTGRP21    type: str22    label: Treatment Group23    derivation: DM.TRTGRP2425  # Qualified aggregates (EX.x) reduce the subject's EX records, matched on26  # USUBJID, the output key EX also carries. A subject with no EX record27  # gets missing from every one of them, counts included.28  #29  # Qualified: counts dose records, including those with a missing dose.30  - name: NDOSE31    type: int32    label: Number of Dose Records33    derivation:34      aggregate:35        expr: "COUNT(EX.*)"3637  # Qualified: counts only the records whose dose is present.38  - name: NDOSEVAL39    type: int40    label: Number of Non-missing Doses41    derivation:42      aggregate:43        expr: "COUNT(EX.EXDOSE)"4445  # Qualified: a missing dose contributes nothing, never zero; a subject46  # whose doses are all missing gets a missing total.47  - name: TOTDOSE48    type: float49    label: Total Dose50    derivation:51      aggregate:52        expr: "SUM(EX.EXDOSE)"5354  # Qualified: defined as SUM over COUNT of the non-missing doses.55  - name: MEANDOSE56    type: float57    label: Mean Dose58    derivation:59      aggregate:60        expr: "MEAN(EX.EXDOSE)"6162  # Qualified: the largest present dose.63  - name: MAXDOSE64    type: float65    label: Maximum Dose66    derivation:67      aggregate:68        expr: "MAX(EX.EXDOSE)"6970  # Qualified with a filter: only the DRUG records are eligible.71  - name: DRUGDOSE72    type: float73    label: Total Drug Dose74    derivation:75      aggregate:76        filter: "EX.EXTRT = 'DRUG'"77        expr: "SUM(EX.EXDOSE)"7879  # Qualified with a filter matching nothing for some subjects: an empty80  # group reads as missing, not zero.81  - name: PLACDOSE82    type: float83    label: Total Placebo Dose84    derivation:85      aggregate:86        filter: "EX.EXTRT = 'PLACEBO'"87        expr: "SUM(EX.EXDOSE)"8889  # Qualified with a derive step: each record's dose is converted to90  # milligrams before the reduction. Bindings evaluate in order; the91  # second reads the value bound by the first.92  - name: TOTDOSE_MG93    type: float94    label: Total Dose in Milligrams95    derivation:96      aggregate:97        derive:98          - name: DOSE99            type: float100            derivation: EX.EXDOSE101          - name: DOSEMG102            type: float103            derivation:104              compute:105                expr: "DOSE * 1000"106        expr: "SUM(DOSEMG)"107108  # Unqualified: reduces the built rows within each treatment group and109  # broadcasts the result to every row in the group. Subject 04 has no110  # dose records, so its NDOSE is missing and excluded from the mean.111  - name: AVGNDOSE112    type: float113    label: Mean Dose Records per Treatment Group114    derivation:115      aggregate:116        expr: "MEAN(NDOSE)"117        group_by: [TRTGRP]

Inputs

Under review2 source files

input/dm.csv

Edit4 rows

input/dm.csv
USUBJID	TRTGRP
	A
	A
	B
	B

input/ex.csv

Edit6 rows

input/ex.csv
EXSEQ	EXDOSE	EXTRT
1	100	DRUG
2	200	DRUG
3	—	DRUG
1	150	DRUG
2	150	PLACEBO
1	—	DRUG

Expected output

Under review4 expected rows

expected/adsl.csv

Edit4 rows

expected/adsl.csv
TRTGRP	NDOSE	NDOSEVAL	TOTDOSE	MEANDOSE	MAXDOSE	DRUGDOSE	PLACDOSE	TOTDOSE_MG	AVGNDOSE
A	3	2	300	150	200	300	—	300000	2.5
A	2	2	300	150	150	150	150	300000	2.5
B	1	0	—	—	—	—	—	—	1
B	—	—	—	—	—	—	—	—	1

Shaded columns are derived— = missing value

Comments and assessment materials

Dataset Name	Variable Name	Variable Label	Type	Computational Method	Origin
ADSL	USUBJID	Unique Subject Identifier	character	Copy value from DM.USUBJID.	Predecessor
ADSL	TRTGRP	Treatment Group	character	Copy value from DM.TRTGRP.	Predecessor
ADSL	NDOSE	Number of Dose Records	numeric	Aggregate COUNT(EX.*) over records matching on the shared output keys.	Derived
ADSL	NDOSEVAL	Number of Non-missing Doses	numeric	Aggregate COUNT(EX.EXDOSE) over records matching on the shared output keys.	Derived
ADSL	TOTDOSE	Total Dose	numeric	Aggregate SUM(EX.EXDOSE) over records matching on the shared output keys.	Derived
ADSL	MEANDOSE	Mean Dose	numeric	Aggregate MEAN(EX.EXDOSE) over records matching on the shared output keys.	Derived
ADSL	MAXDOSE	Maximum Dose	numeric	Aggregate MAX(EX.EXDOSE) over records matching on the shared output keys.	Derived
ADSL	DRUGDOSE	Total Drug Dose	numeric	Aggregate SUM(EX.EXDOSE) over records matching on the shared output keys, where EX.EXTRT = 'DRUG'.	Derived
ADSL	PLACDOSE	Total Placebo Dose	numeric	Aggregate SUM(EX.EXDOSE) over records matching on the shared output keys, where EX.EXTRT = 'PLACEBO'.	Derived
ADSL	TOTDOSE_MG	Total Dose in Milligrams	numeric	Aggregate SUM(DOSEMG) over records matching on the shared output keys, with DOSE as EX.EXDOSE, with DOSEMG as compute DOSE * 1000.	Derived
ADSL	AVGNDOSE	Mean Dose Records per Treatment Group	numeric	Aggregate MEAN(NDOSE) over records grouped by TRTGRP, matching on the shared output keys.	Derived

Version	Date	Author	Description
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

run.R

Edit4 lines

1library(yamaa)23adsl <- yamaa_domain("spec.yaml")$output4adsl

run.py

Edit4 lines

1import yamaa23adsl = yamaa.yamaa_domain("spec.yaml").output4adsl

input/dm.csv

input/ex.csv

expected/adsl.csv

Comments

Mapping spec

yamaa code

run.R

run.py