# Movement Thinking: primary-study methods package

Version 3.0 · 15 October 2026 · The Research University

This package accompanies **From Instructional Leadership to the Design of Learning Work: A Movement Thinking Framework and Empirical Evaluation Strategy for Future-Ready Schools**, by Dr. Freedom Cheteni.

It contains a proposed primary-study protocol, original draft instruments, a completed SEDA file audit, and simulation-based design calculations. It contains no local participant observations, completed primary trial, fitted mediation model, ethics approval or trial registration. Stanford and OECD are credited data providers; the package does not establish a research partnership or advisory appointment.

## Files

| File | Purpose |
| --- | --- |
| study-specification.json | Versioned design, estimand, endpoints and activation prerequisites |
| instruments.md | Original assessment rubric, fictional task packet and implementation measures |
| analysis-specification.md | Primary estimator, missing-data model, inference and mechanism boundaries |
| source-audit.json | URLs, document/file hashes, access dates and verified variable mappings |
| seda-feasibility-audit.json | Executed public-file counts and context-slice checks |
| design-analysis.json | Executed artificial-data simulation; assumptions, seeds and Monte Carlo error |
| design-sensitivity.png / plot_design.py | Figure and regeneration code using design-analysis.json |
| research_tools.py | Allocation, complete-cohort analysis, design simulation, SEDA audit and a restricted TALIS BRR estimator |
| requirements.txt | Versions used for the executed calculations |

## Reproduce the completed analyses

Use Python 3.12 or a compatible Python 3 environment with the dependencies in requirements.txt. Install in a dedicated environment, retaining the Python and package versions with the outputs.

```sh
python -m pip install -r requirements.txt
python research_tools.py design --seed 20261015 --replicates 10000 --output design-analysis.json
python plot_design.py
python research_tools.py audit-seda --input /private-data/seda_admindist_pool_cs_2025.2.csv --output local-seda-audit.json
```

Obtain the SEDA file from its original Stanford repository link in source-audit.json and follow the provider's data-use conditions. The original CSV is not redistributed here. Check its SHA-256 against the published audit before comparing counts. A revised provider file requires a new provenance record and audit rather than silently replacing the input under an old version label.

The simulation reports a model-based rejection frequency for an assumed standardized effect. It does not estimate a Movement Thinking effect. Its generated normal draws are artificial and must never be described as anonymized student data. The SEDA audit reports source-file structure and availability, not changes caused by the proposed intervention.

## Tools for a future study

```sh
python research_tools.py randomize --input /private-data/locked-school-roster.json --seed 123456 --output /private-data/allocation.json
python research_tools.py analyze-complete --input /private-data/complete-cohort.json --permutations 9999 --output /private-data/analysis.json
```

The seed above is an example command argument, not the allocation seed for a real study. An independent custodian must generate and preserve the actual allocation after the cohort and baseline records are locked. A roster has `data_kind` and `schools`; every school has globally unique `school_id` and `district_id`. Every district block must contain exactly four schools. The command assigns two to each condition.

The complete-cohort input additionally has `protocol_version: "3.0"`, `students`, and each school's binary `assignment`. Each student record contains a unique `student_id`, `school_id`, `baseline_score` and `endline_score`. Scores are finite numbers in 0–20, and there are exactly 40 baseline-cohort records per school. `data_kind` must explicitly distinguish `synthetic_demonstration` from `approved_study_data`. This is investigator-supplied provenance, not software verification of ethical approval.

The adapter rejects incomplete cohorts and missing scores. It does not delete incomplete records, impute outcomes, implement the full missing-data specification, or fit an SEM. Investigators must implement and independently validate the separate missing-data plan before study activation. Keep participant inputs and outputs outside the public website and repository. A successful script execution is not evidence that the study has the necessary permissions or that its model assumptions hold.

## TALIS estimator

```sh
python research_tools.py talis-brr --input /private-data/teacher-release.csv --country USA --variable TT4G26G --category 6 --output /private-data/talis-result.json
```

This narrowly scoped estimator requires the actual OECD teacher file, country/population fields, TCHWGT and all 100 teacher replicate weights. It estimates a valid-response category proportion in ISCED level 2, using Fay factor 0.5. It is not an SEM implementation or a validated Movement Thinking score. It has been tested with artificial fixtures, not with a downloaded TALIS respondent file. Verify release-specific availability, special values, national adaptations and participation qualifications; the U.S. sample has important adjudication cautions in the official guide.

There is no public TALIS-to-SEDA join operation. The public TALIS identifiers are scrambled. A future authorized linkage would need an approved crosswalk and a compatible study population. An identical-looking number is not an authorization or a shared identifier.

## Version and review record

The preceding conceptual edition and its publisher-confirmed review by Dr. Vickie Aultmon are preserved separately. Version 3.0 adds the primary-study protocol, executed public-file audit and methods code. Review attribution for the earlier text does not automatically cover these substantive additions. Editorial decisions, author approval, external registration and a DOI remain distinct records.

The original task and rubric are draft research instruments, not a validated commercial assessment. Source papers and data retain their original authors, institutions, dates and rights. Consult the article's references and attribution statement before reusing third-party material.