Mean NPI-X Total Score from Week 4 through Week 24 - Windowed

The secondary endpoint — mean NPI-X total score across the Weeks 4 to 24 window. The one display where the study's analysis metadata and its own published report disagree about which score to use, and the footnotes say which side was taken.

Slug: t-eff-npix-mean
Regulatory ID: EFT09
Type: Table
E3 position: §14.2
Status: evidenced
Current iteration: v003
Datasets: adqsnpix a8db42f, adsl 70c7278
Evidence: 26/26 passing
Requirements: DSP-ALL-001 DSP-ALL-002 DSP-EFF-001 DSP-EFF-002 DSP-EFF-003 DSP-EFF-004 DSP-EFF-005 DSP-EFF-006 DSP-EFF-007 DSP-EFF-008 DSP-EFF-009 DSP-INTXT-001 DSP-INTXT-002 DSP-INTXT-003 STD-GATE-001 STD-GATE-002 STD-GATE-003 STD-GATE-004 STD-GATE-005 STD-MODEL-001 STD-MODEL-002 STD-MODEL-003 STD-MODEL-004 STD-MODEL-005 STD-SRC-001 STD-SRC-002
Used in: Clinical Study Report Table 14.2.12
Post-Text Display Package Table 14.2.12
Abbreviated Clinical Study Report (ICH E3) Table 14.2.12
Study model: Efficacy Analysis Set EFFFL · grouped by TRTP
Specification: 3 iterations
Environment: R 4.3.3 Darwin 23.6.0
Bound by: No text block binds it
Values: No value is sourced from it

Inputs

adsl254 rows · 70c7278
adqsnpix31140 rows · a8db42f
analysis.yamlwhat to compute
display.yamlhow to show it
custom.Rthe display's own method
study modelEfficacy Analysis Set

Function

prepare_data()regenerate('t-eff-npix-mean')
build_ard()one row per statistic
render_display()every variant
render_rtf()the same cells

Outputs

ard.jsonv003 · 71 rows · 670210c
table.htmlthe rendered display
table.rtfsha256 51085b2
manifest.jsonwho, why, every hash

RTFFull display51085b2

Generated by the same pipeline run that wrote the ARD and the HTML above, from the same rendered cells; the short hash is the manifest's record of the file.

t-eff-npix-mean · post_text variant · study CDISCPILOT01 · data cut-off 2015-02-17

	Placebo (N=79)	Xanomeline Low Dose (N=81)	Xanomeline High Dose (N=74)
Mean NPI-X Total Score from Week 4 through Week 24 - Windowed
Study CDISCPILOT01 — Efficacy Analysis Set
Baseline
n	79	81	74
Mean (SD)	9.5 (12.10)	8.7 (9.82)	11.9 (13.70)
Median (Range)	5.0 (0;66)	4.0 (0;32)	8.0 (0;61)
Mean of Weeks 4-24
n	78	75	69
Mean (SD)	9.3 (11.18)	9.1 (12.10)	9.6 (11.60)
Median (Range)	5.5 (0;65)	3.8 (0;51)	4.4 (0;46)
Statistical comparison of the mean over Weeks 4 to 24
p-value (dose response) [1][2]			0.637
p-value (Xanomeline - Placebo) [1][3]		0.760	0.517
Difference of LS means (SE)		0.3 (1.13)	-0.7 (1.15)
95% CI		(-1.9;2.6)	(-3.0;1.5)
p-value (Xanomeline High - Xanomeline Low) [1][3]			0.350
Difference of LS means (SE)			-1.1 (1.17)
95% CI			(-3.4;1.2)
[1] Based on an analysis of covariance (ANCOVA) model with treatment and site group as factors and the baseline score as a covariate.
[2] Test for a non-zero coefficient for treatment (dose) as a continuous variable.
[3] Pairwise comparison with treatment as a categorical variable; p-values are not adjusted for multiple comparisons.
The Week 4 to Week 24 endpoint is the mean of a subject's available NPI-X (9) total scores over the Week 4 to Week 24 assessment windows, as the statistical analysis plan defines it in section 10.2.1. A subject contributes one value however many assessments it averages, and a subject with no assessment in that span contributes none, so n is smaller than the column N.
DEFINED HERE, NOT INHERITED: this endpoint is derived from the per-visit NPTOT records rather than read from the study's own NPTOTMN parameter, because the study's two authoritative statements about this table disagree. Its analysis results metadata (define.xml, Table_14-3.12) selects PARAMCD NPTOTMN, which covers 210 of the 222 efficacy subjects who have an assessment in the Week 4 to Week 24 span: it omits 12 — eleven contributing a single assessment and one contributing two — and adds none. Selecting it reports n = 76, 69, 65. Its own reference report reports n = 78, 75, 69, which is every subject with an assessment in the span, and which the analysis plan's wording ('the mean of all available total scores between Weeks 4 and 24, inclusive') describes. The derivation used here follows the analysis plan and reproduces the reference report to every cell, including the ANCOVA. The divergence is measured and recorded in quality/data/efficacy-agreement.json.
Assessments are selected by AVISITN, not AVISIT: adqsnpix ships AVISIT right-aligned in a 16-character field, so a comparison against the visible label silently selects no records.
Treatment columns are planned treatment (TRTP), which is how the CDISCPILOT01 reference report presents them.
CDISCPILOT01 states no data cut-off date. The date shown is the latest assessment date present in the vendored ADaM package.
Source: adqsnpix (NPI-X (9) total score, PARAMCD NPTOT) and adsl, from the CDISC pilot's own ADaM package (phuse-org/phuse-scripts, MIT). Analysis as specified in the CDISCPILOT01 statistical analysis plan, section 10.2.1. Reference report display: Table 14-3.12. Data cut-off: 2015-02-17.
open.csr display t-eff-npix-mean (post_text variant); generated from the committed ARD.

Rendered from outputs/t-eff-npix-mean/v003/table.html.

analysis	group1	group1_level	group2	group2_level	variable	variable_level	context	stat_name	stat_label	warning	error	stat
population	TRTP	Placebo	—	—	USUBJID	—	population	N	N	—	—	79
population	TRTP	Xanomeline High Dose	—	—	USUBJID	—	population	N	N	—	—	74
population	TRTP	Xanomeline Low Dose	—	—	USUBJID	—	population	N	N	—	—	81
baseline	TRTP	Placebo	—	—	AVAL	—	continuous	N	N	—	—	79
baseline	TRTP	Placebo	—	—	AVAL	—	continuous	mean	Mean	—	—	9.45569620253164
baseline	TRTP	Placebo	—	—	AVAL	—	continuous	sd	SD	—	—	12.1029742777576
baseline	TRTP	Placebo	—	—	AVAL	—	continuous	median	Median	—	—	5
baseline	TRTP	Placebo	—	—	AVAL	—	continuous	p25	Q1	—	—	1
baseline	TRTP	Placebo	—	—	AVAL	—	continuous	p75	Q3	—	—	12
baseline	TRTP	Placebo	—	—	AVAL	—	continuous	min	Min	—	—	0
baseline	TRTP	Placebo	—	—	AVAL	—	continuous	max	Max	—	—	66
baseline	TRTP	Xanomeline High Dose	—	—	AVAL	—	continuous	N	N	—	—	74
baseline	TRTP	Xanomeline High Dose	—	—	AVAL	—	continuous	mean	Mean	—	—	11.8918918918919
baseline	TRTP	Xanomeline High Dose	—	—	AVAL	—	continuous	sd	SD	—	—	13.7038816294574
baseline	TRTP	Xanomeline High Dose	—	—	AVAL	—	continuous	median	Median	—	—	8
baseline	TRTP	Xanomeline High Dose	—	—	AVAL	—	continuous	p25	Q1	—	—	2
baseline	TRTP	Xanomeline High Dose	—	—	AVAL	—	continuous	p75	Q3	—	—	17
baseline	TRTP	Xanomeline High Dose	—	—	AVAL	—	continuous	min	Min	—	—	0
baseline	TRTP	Xanomeline High Dose	—	—	AVAL	—	continuous	max	Max	—	—	61
baseline	TRTP	Xanomeline Low Dose	—	—	AVAL	—	continuous	N	N	—	—	81
baseline	TRTP	Xanomeline Low Dose	—	—	AVAL	—	continuous	mean	Mean	—	—	8.7283950617284
baseline	TRTP	Xanomeline Low Dose	—	—	AVAL	—	continuous	sd	SD	—	—	9.82472944370354
baseline	TRTP	Xanomeline Low Dose	—	—	AVAL	—	continuous	median	Median	—	—	4
baseline	TRTP	Xanomeline Low Dose	—	—	AVAL	—	continuous	p25	Q1	—	—	1
baseline	TRTP	Xanomeline Low Dose	—	—	AVAL	—	continuous	p75	Q3	—	—	15
baseline	TRTP	Xanomeline Low Dose	—	—	AVAL	—	continuous	min	Min	—	—	0
baseline	TRTP	Xanomeline Low Dose	—	—	AVAL	—	continuous	max	Max	—	—	32
mean_4_24	TRTP	Placebo	—	—	AVAL	—	continuous	N	N	—	—	78
mean_4_24	TRTP	Placebo	—	—	AVAL	—	continuous	mean	Mean	—	—	9.31155141155141
mean_4_24	TRTP	Placebo	—	—	AVAL	—	continuous	sd	SD	—	—	11.1812314549378
mean_4_24	TRTP	Placebo	—	—	AVAL	—	continuous	median	Median	—	—	5.50505050505051
mean_4_24	TRTP	Placebo	—	—	AVAL	—	continuous	p25	Q1	—	—	1.27272727272727
mean_4_24	TRTP	Placebo	—	—	AVAL	—	continuous	p75	Q3	—	—	12.4545454545455
mean_4_24	TRTP	Placebo	—	—	AVAL	—	continuous	min	Min	—	—	0
mean_4_24	TRTP	Placebo	—	—	AVAL	—	continuous	max	Max	—	—	64.5
mean_4_24	TRTP	Xanomeline High Dose	—	—	AVAL	—	continuous	N	N	—	—	69
mean_4_24	TRTP	Xanomeline High Dose	—	—	AVAL	—	continuous	mean	Mean	—	—	9.58007037246168
mean_4_24	TRTP	Xanomeline High Dose	—	—	AVAL	—	continuous	sd	SD	—	—	11.6018032718414
mean_4_24	TRTP	Xanomeline High Dose	—	—	AVAL	—	continuous	median	Median	—	—	4.36363636363636
mean_4_24	TRTP	Xanomeline High Dose	—	—	AVAL	—	continuous	p25	Q1	—	—	0.3
mean_4_24	TRTP	Xanomeline High Dose	—	—	AVAL	—	continuous	p75	Q3	—	—	16
mean_4_24	TRTP	Xanomeline High Dose	—	—	AVAL	—	continuous	min	Min	—	—	0
mean_4_24	TRTP	Xanomeline High Dose	—	—	AVAL	—	continuous	max	Max	—	—	46.2727272727273
mean_4_24	TRTP	Xanomeline Low Dose	—	—	AVAL	—	continuous	N	N	—	—	75
mean_4_24	TRTP	Xanomeline Low Dose	—	—	AVAL	—	continuous	mean	Mean	—	—	9.06895412457913
mean_4_24	TRTP	Xanomeline Low Dose	—	—	AVAL	—	continuous	sd	SD	—	—	12.0993154805808
mean_4_24	TRTP	Xanomeline Low Dose	—	—	AVAL	—	continuous	median	Median	—	—	3.77777777777778
mean_4_24	TRTP	Xanomeline Low Dose	—	—	AVAL	—	continuous	p25	Q1	—	—	0.75
mean_4_24	TRTP	Xanomeline Low Dose	—	—	AVAL	—	continuous	p75	Q3	—	—	15.2857142857143
mean_4_24	TRTP	Xanomeline Low Dose	—	—	AVAL	—	continuous	min	Min	—	—	0
mean_4_24	TRTP	Xanomeline Low Dose	—	—	AVAL	—	continuous	max	Max	—	—	51
ancova	TRTP	Xanomeline High Dose	—	—	AVAL	dose_response	ancova	pvalue	pvalue	—	—	0.637355911747683
ancova	TRTP	Xanomeline High Dose	—	—	AVAL	dose_response	ancova	n_model	n_model	—	—	222
ancova	TRTP	Xanomeline Low Dose	—	—	AVAL	xan_vs_placebo	ancova	estimate	estimate	—	—	0.344970498781485
ancova	TRTP	Xanomeline Low Dose	—	—	AVAL	xan_vs_placebo	ancova	se	se	—	—	1.12595262472192
ancova	TRTP	Xanomeline Low Dose	—	—	AVAL	xan_vs_placebo	ancova	pvalue	pvalue	—	—	0.759621031087313
ancova	TRTP	Xanomeline Low Dose	—	—	AVAL	xan_vs_placebo	ancova	lcl	lcl	—	—	-1.87477152944715
ancova	TRTP	Xanomeline Low Dose	—	—	AVAL	xan_vs_placebo	ancova	ucl	ucl	—	—	2.56471252701012
ancova	TRTP	Xanomeline Low Dose	—	—	AVAL	xan_vs_placebo	ancova	n_model	n_model	—	—	222
ancova	TRTP	Xanomeline High Dose	—	—	AVAL	xan_vs_placebo	ancova	estimate	estimate	—	—	-0.748808650896317
ancova	TRTP	Xanomeline High Dose	—	—	AVAL	xan_vs_placebo	ancova	se	se	—	—	1.15337540534439
ancova	TRTP	Xanomeline High Dose	—	—	AVAL	xan_vs_placebo	ancova	pvalue	pvalue	—	—	0.51690443294829
ancova	TRTP	Xanomeline High Dose	—	—	AVAL	xan_vs_placebo	ancova	lcl	lcl	—	—	-3.02261289927708
ancova	TRTP	Xanomeline High Dose	—	—	AVAL	xan_vs_placebo	ancova	ucl	ucl	—	—	1.52499559748445
ancova	TRTP	Xanomeline High Dose	—	—	AVAL	xan_vs_placebo	ancova	n_model	n_model	—	—	222
ancova	TRTP	Xanomeline High Dose	—	—	AVAL	high_vs_low	ancova	estimate	estimate	—	—	-1.0937791496778
ancova	TRTP	Xanomeline High Dose	—	—	AVAL	high_vs_low	ancova	se	se	—	—	1.16689533939629
ancova	TRTP	Xanomeline High Dose	—	—	AVAL	high_vs_low	ancova	pvalue	pvalue	—	—	0.349670131390298
ancova	TRTP	Xanomeline High Dose	—	—	AVAL	high_vs_low	ancova	lcl	lcl	—	—	-3.39423706462939
ancova	TRTP	Xanomeline High Dose	—	—	AVAL	high_vs_low	ancova	ucl	ucl	—	—	1.20667876527378
ancova	TRTP	Xanomeline High Dose	—	—	AVAL	high_vs_low	ancova	n_model	n_model	—	—	222

71 rows, one per computed statistic.

Schema: opencsr/ard/v1
Generated: 2026-09-02T15:23:25Z
Spec hash: sha256:80f5f0ccff24b1cf2f942ea6b2e3615b2fe5bf61016f79e950b6c75d4df5019f
Display hash: sha256:c3210952b525d9075afa6c8d2ecf1772a52200499d0202ac66a23888400bcbc4
ARD file: outputs/t-eff-npix-mean/v003/ard.json
ARD hash: sha256:670210cf013360c0acf6a92ee1315b14c3bb0085b5afd2774e45f299bdb00d5c

Source datasets

Dataset	Rows	Package	Hash
adsl	254	phuse-org/phuse-scripts:data/adam 398a6d33ced9359ffb58c46650a6d488811176b1	sha256:70c7278e83e571a7d4a94fd864c98bf8ca243c7c2cc9eec8b99a40fa7bc98e35
adqsnpix	31140	phuse-org/phuse-scripts:data/adam 398a6d33ced9359ffb58c46650a6d488811176b1	sha256:a8db42f422294f16537326a268fbc7d21e54e50677f628ad868849f4d7376971

Environment

R 4.3.3 · Darwin 23.6.0 · cards 0.6.1 · cardx 0.2.5 · gtsummary 2.3.0 · gt 1.0.0 · dplyr 1.1.4 · pharmaverseadam 1.1.0 · jsonlite 2.0.0 · yaml 2.3.10 · digest 0.6.37

library/tfl/t-eff-npix-mean/analysis.yaml

id: t-eff-npix-mean
title: "Mean NPI-X Total Score from Week 4 through Week 24 - Windowed"
regulatory_id: EFT09
type: table
dataset: adqsnpix
analysis_set: efficacy
denominator: adsl
sources: phuse
group:
  - TRTP
total: false
# Visits are selected by AVISITN, never by AVISIT. ADQSNPIX ships AVISIT
# right-aligned into a 16-character field ("      Weeks 4-24"), so the obvious
# `AVISIT == 'Weeks 4-24'` selects nothing at all while looking correct. AVISITN
# carries the same fact as a number: 0 baseline, 2..26 the scheduled weeks, 98
# the study's own derived Weeks 4-24 average.
analyses:
  - name: population
    method: continuous
    custom: population_n
    label: "Subjects in the analysis population"
    filter: "PARAMCD == 'NPTOT' & AVISITN == 0"

  - name: baseline
    method: continuous
    label: "NPI-X (9) total score at baseline"
    variables: [AVAL]
    filter: "PARAMCD == 'NPTOT' & AVISITN == 0"

  # The SAP defines this endpoint as "the mean of all available total scores
  # between Weeks 4 and 24, inclusive" (section 10.2.1) — one number per
  # subject, derived here rather than read from the study's own NPTOTMN
  # parameter. See the display's source note for why.
  - name: mean_4_24
    method: continuous
    custom: subject_mean
    label: "Per-subject mean NPI-X (9) total score over Weeks 4 to 24"
    variables: [AVAL]
    filter: "PARAMCD == 'NPTOT' & ANL01FL == 'Y' & AVISITN >= 4 & AVISITN <= 24"
    derive:
      statistic: mean
      value: AVAL
      id: USUBJID
      carry: [TRTPN, SITEGR1, BASE]

  - name: ancova
    method: continuous
    custom: subject_mean_ancova
    label: "ANCOVA of the per-subject mean over Weeks 4 to 24"
    variables: [AVAL]
    filter: "PARAMCD == 'NPTOT' & ANL01FL == 'Y' & AVISITN >= 4 & AVISITN <= 24"
    derive:
      statistic: mean
      value: AVAL
      id: USUBJID
      carry: [TRTPN, SITEGR1, BASE]
    response: AVAL
    covariates: [BASE]
    factors: [SITEGR1]
    treatment: TRTP
    treatment_levels: ["Placebo", "Xanomeline Low Dose", "Xanomeline High Dose"]
    dose: TRTPN
    conf_level: 0.95
    contrasts:
      - { name: dose_response, type: dose, column: "Xanomeline High Dose" }
      - { name: xan_vs_placebo, type: pairwise, test: "Xanomeline Low Dose",
          reference: "Placebo", column: "Xanomeline Low Dose" }
      - { name: xan_vs_placebo, type: pairwise, test: "Xanomeline High Dose",
          reference: "Placebo", column: "Xanomeline High Dose" }
      - { name: high_vs_low, type: pairwise, test: "Xanomeline High Dose",
          reference: "Xanomeline Low Dose", column: "Xanomeline High Dose" }

library/tfl/t-eff-npix-mean/display.yaml

id: t-eff-npix-mean
title: "Mean NPI-X Total Score from Week 4 through Week 24 - Windowed"
study: CDISCPILOT01
population_label: "Efficacy Analysis Set"
cutoff: "2015-02-17"
source: "Source: adqsnpix (NPI-X (9) total score, PARAMCD NPTOT) and adsl, from the CDISC pilot's own ADaM package (phuse-org/phuse-scripts, MIT). Analysis as specified in the CDISCPILOT01 statistical analysis plan, section 10.2.1. Reference report display: Table 14-3.12. Data cut-off: 2015-02-17."
footnotes:
  - "[1] Based on an analysis of covariance (ANCOVA) model with treatment and site group as factors and the baseline score as a covariate."
  - "[2] Test for a non-zero coefficient for treatment (dose) as a continuous variable."
  - "[3] Pairwise comparison with treatment as a categorical variable; p-values are not adjusted for multiple comparisons."
  - "The Week 4 to Week 24 endpoint is the mean of a subject's available NPI-X (9) total scores over the Week 4 to Week 24 assessment windows, as the statistical analysis plan defines it in section 10.2.1. A subject contributes one value however many assessments it averages, and a subject with no assessment in that span contributes none, so n is smaller than the column N."
  - "DEFINED HERE, NOT INHERITED: this endpoint is derived from the per-visit NPTOT records rather than read from the study's own NPTOTMN parameter, because the study's two authoritative statements about this table disagree. Its analysis results metadata (define.xml, Table_14-3.12) selects PARAMCD NPTOTMN, which covers 210 of the 222 efficacy subjects who have an assessment in the Week 4 to Week 24 span: it omits 12 — eleven contributing a single assessment and one contributing two — and adds none. Selecting it reports n = 76, 69, 65. Its own reference report reports n = 78, 75, 69, which is every subject with an assessment in the span, and which the analysis plan's wording ('the mean of all available total scores between Weeks 4 and 24, inclusive') describes. The derivation used here follows the analysis plan and reproduces the reference report to every cell, including the ANCOVA. The divergence is measured and recorded in quality/data/efficacy-agreement.json."
  - "Assessments are selected by AVISITN, not AVISIT: adqsnpix ships AVISIT right-aligned in a 16-character field, so a comparison against the visible label silently selects no records."
  - "Treatment columns are planned treatment (TRTP), which is how the CDISCPILOT01 reference report presents them."
  - "CDISCPILOT01 states no data cut-off date. The date shown is the latest assessment date present in the vendored ADaM package."
columns:
  order: ["Placebo", "Xanomeline Low Dose", "Xanomeline High Dose"]
format:
  mean_sd: "{mean} ({sd})"
  median_range: "{median} ({min};{max})"
  est_se: "{estimate} ({se})"
  ci: "({lcl};{ucl})"
  p_value: "{pvalue}"
  digits:
    mean: 1
    sd: 2
    median: 1
    min: 0
    max: 0
    estimate: 1
    se: 2
    lcl: 1
    ucl: 1
    pvalue: 3
rows:
  - { label: "Baseline", section: true }
  - { analysis: baseline, variable: AVAL, pattern: "N", label: "n", indent: 1 }
  - { analysis: baseline, variable: AVAL, pattern: mean_sd, label: "Mean (SD)", indent: 1 }
  - { analysis: baseline, variable: AVAL, pattern: median_range, label: "Median (Range)", indent: 1 }
  - { label: "Mean of Weeks 4-24", section: true }
  - { analysis: mean_4_24, variable: AVAL, pattern: "N", label: "n", indent: 1 }
  - { analysis: mean_4_24, variable: AVAL, pattern: mean_sd, label: "Mean (SD)", indent: 1 }
  - { analysis: mean_4_24, variable: AVAL, pattern: median_range, label: "Median (Range)", indent: 1 }
  - { label: "Statistical comparison of the mean over Weeks 4 to 24", section: true }
  - { analysis: ancova, variable: AVAL, level: dose_response, pattern: p_value,
      label: "p-value (dose response) [1][2]", indent: 1 }
  - { analysis: ancova, variable: AVAL, level: xan_vs_placebo, pattern: p_value,
      label: "p-value (Xanomeline - Placebo) [1][3]", indent: 1 }
  - { analysis: ancova, variable: AVAL, level: xan_vs_placebo, pattern: est_se,
      label: "Difference of LS means (SE)", indent: 2 }
  - { analysis: ancova, variable: AVAL, level: xan_vs_placebo, pattern: ci,
      label: "95% CI", indent: 2 }
  - { analysis: ancova, variable: AVAL, level: high_vs_low, pattern: p_value,
      label: "p-value (Xanomeline High - Xanomeline Low) [1][3]", indent: 1 }
  - { analysis: ancova, variable: AVAL, level: high_vs_low, pattern: est_se,
      label: "Difference of LS means (SE)", indent: 2 }
  - { analysis: ancova, variable: AVAL, level: high_vs_low, pattern: ci,
      label: "95% CI", indent: 2 }
variants:
  post_text: {}

library/tfl/t-eff-npix-mean/custom.R

# Display-specific statistics (contract section 2).
#
# Thin wrappers, as the ADAS-Cog displays are: the statistics live in {opencsr}
# so they are tested once and shared, and every model term and every derivation
# rule is declared in this display's analysis.yaml rather than written here.
#
# The secondary efficacy endpoint is a statistic of a subject, not of a record —
# the mean of that subject's available total scores over the Week 4 to Week 24
# windows — so both functions collapse the records to one per subject before
# anything is computed. `derive_subject_summary()` refuses to carry a column
# that varies inside a subject, which is what stops the covariate in the model
# below from silently depending on record order.

subject_mean <- function(data, spec, denominator = NULL) {
  opencsr::ard_derived_continuous(data, spec, denominator)
}

subject_mean_ancova <- function(data, spec, denominator = NULL) {
  opencsr::ard_derived_ancova(data, spec, denominator)
}

population_n <- function(data, spec, denominator = NULL) {
  opencsr::ard_population_n(data, spec, denominator)
}

Every iteration is kept: outputs/t-eff-npix-mean/vNNN/ holds the spec snapshot, the ARD, the rendered output, and a manifest recording who regenerated it, why, and from which commit.