# A10 · Community-governed cultural authority trial

**Who May Change the Learning Place? A Community-Governed Trial of Cultural Authority, School Belonging and Disciplinary Transfer**

Dr. Freedom Cheteni · Journal for School Superintendents · The Research University · version 1.0 · verified 16 October 2026 UTC.

This new primary-study protocol accompanies the preserved conceptual article *Belonging by Design*. Completed work consists of a real administrative-data feasibility audit and artificial design calculations. No participants, intervention effects, ethics approval, registration, reviewer, acceptance or article DOI are represented as completed.

## What was executed

The official NCES CCD 2023–24 version 1a directory contains **102,274 rows and unique school identifiers**. The declared active regular grade-7/8 frame yields **29,115 schools**; **16,941** are in **3,049 LEAs** with at least two candidates. These are historical administrative possibilities, not current recruitment or permissions.

The membership CSV contains **11,209,338 long rows and 99,678 schools**. Restricting `TOTAL_INDICATOR` to `Education Unit Total` yields 99,678 unique school totals, with no unmatched school IDs. Across the **full 11,209,338-row long membership file**, there are **66,138 blank student counts**: 12,421 flagged Missing, 53,567 Not reported and 150 Suppressed. These are detailed-cell counts across all record types, not the number of unavailable school totals. They remain unavailable, never zero. The selected total may include adult education and does not establish 40 eligible grade-7/8 students. Race/ethnicity reporting categories do not identify culture or authorize representations. Full detailed-cell uniqueness was not checked; selected school-total uniqueness was.

Twelve artificial scenarios examine 24, 32 or 40 pairs, ICC .05 or .10 and comparison exposure 0 or .25, with **10,000 replicates per scenario**. Every school starts with 40 students; retained sizes stay within 12–40. At 32 pairs, power ranges from **77.84% to 99.58%** under declared assumptions. These are planning results, not field effects. The simulation does not implement or validate the proposed imputation or ordinal-measurement models.

## Files

- `article.md`: complete manuscript and references.
- `study-specification.json`: proposed population, estimand, governance and activation requirements.
- `instruments.md`: original transfer rubric and governance forms; identifies the proposed PSSM without reproducing copyrighted items or granting permissions.
- `analysis-specification.md`: contrasts, missingness, measurement, interference and simulation details.
- `ccd-feasibility-audit.json`: executed administrative diagnostics.
- `design-analysis.json`: all 12 scenarios, achieved retention, power, null rejection and Monte Carlo error.
- `research_tools.py`: executable local methods; no network calls or participant data.
- `test_methods.py` and `test-results.json`: 12 synthetic-fixture/methods checks and recorded results.
- `source-audit.json`: exact authoritative URLs, SHA-256 hashes, verification date, access limits and reference-specific evidence.
- `requirements.txt`: actual NumPy, SciPy and Matplotlib versions used.
- `plot_design.py` and `design-sensitivity.png`: reproducible scientific plot of artificial results.

## Reproduce the completed calculations

Use Python 3.11 or newer in an isolated environment. Install the pinned requirements. Download the official source archives to a local data directory using the URLs in `source-audit.json`, and verify their hashes. Do not upload participant records or raw archives to the journal.

```bash
python research_tools.py audit --directory /path/ccd_sch_029_2324_w_1a_073124.zip --membership /path/ccd_SCH_052_2324_l_1a_073124.zip --output repeated-ccd-audit.json
python research_tools.py design --seed 20261016 --replicates 10000 --output repeated-design.json
python test_methods.py
```

The membership CSV expands to approximately 2.35 GB. The streaming audit does not extract or retain the entire file in memory. It processes all rows but checks duplicates only for the selected school-total join, not the full detailed-cell key. Tests use synthetic fixtures and committed aggregate diagnostics. They do not certify ethics or data collection. The pair-summary utility verifies complete-data mathematics; it is not a full trial-analysis engine.

## Before primary research

Finalize instrument permissions and the independent measurement pilot, local community/rightsholder governance and material permissions, safeguarding and ethics determination, compensation and accessibility budgets, validated analysis implementation and prospective registration. Confirm contemporary eligibility and shared networks before allocation.

Students may use private or fictional tasks. No personal cultural story, public profile, ORCID account, publication or AI-training contribution is required. No cultural knowledge, student record, recruitment list, rights-bearing contribution or copyrighted scale item is included. Citation does not grant reuse permission. A paper companion can retrieve passages; it cannot act as a participant, cultural representative, human reviewer or consent authority.