EN
EnglishSimplified ChineseArabic
OPEN RESEARCH EDITIONFramework 1.0Academic Revision 1.2
FRAMEWORK 1.0 / Methodology

Methodology

A claim enters the research record only through an explicit, versioned and reproducible protocol.

Authoritative English editionDownload the authoritative English PDF
01

Minimum empirical workflow

  1. 01

    Register the question

    State the claim, target population, system boundary, scale, time window, primary outcome, and evidence that would count against the claim.

  2. 02

    Map prior literature

    Identify established findings, competing theories, unresolved gaps, and domain-specific baselines before presenting an original construct.

  3. 03

    Operationalize constructs

    Define every construct in plain language; specify indicators, exclusions, unit, transformations, measurement error, and expected bias.

  4. 04

    Reduce dimensions explicitly

    Translate the multilevel ontology into a minimal identifiable state vector while preserving network, distributional, and unequal-power information that matters.

  5. 05

    Freeze the data plan

    Register sources, provenance, inclusion rules, missingness, harmonization, licenses, privacy controls, and the data cutoff before outcome analysis.

  6. 06

    Specify the measurement model

    Declare whether constructs are reflective, formative, network-based, latent, or composite; test reliability, validity, and invariance.

  7. 07

    Specify competing models

    Compare the canonical multiplicative relation with additive, latent-variable, domain-specific, causal, and regularized machine-learning alternatives.

  8. 08

    Estimate and calibrate

    Separate training, calibration, validation, and prospective evaluation; report priors, parameters, uncertainty, convergence, and diagnostics.

  9. 09

    Backtest without leakage

    Use historical cutoffs, preregistered events and horizons, frozen vintages, naive baselines, false alarms, missed events, and lead-time evaluation.

  10. 10

    Stress-test conclusions

    Vary proxies, boundaries, weights, ε, links, lags, missing-data rules, dimensional-reduction choices, and influential cases.

  11. 11

    Replicate independently

    Release lawful data, code, environments, decision logs, and persistent versions; invite an independent team before upgrading claim status.

  12. 12

    Report use limits

    Publish negative results, subgroup error, ethical constraints, dual-use risk, abstention conditions, model drift, and the conditions requiring revision or withdrawal.

02

Evidence maturity gates

01

G0 · Proposal

Question, scope, ownership, and falsification condition are public.

02

G1 · Operationalized

Construct dictionary, measurement model, and pilot data exist.

03

G2 · Internally tested

Code, diagnostics, uncertainty, and negative results are recorded.

04

G3 · Externally replicated

An independent team reproduces or challenges the result.

05

G4 · Prospectively validated

Performance is demonstrated on unseen future data.

06

G5 · Practice-qualified

Benefits, harms, rights, drift, and governance are evaluated in use.