Skip to content

version-roadmap.md

Latest commit

 

History

History
111 lines (74 loc) · 4.27 KB

version-roadmap.md

File metadata and controls

111 lines (74 loc) · 4.27 KB
Raw

Version Roadmap

EvidenceForge uses early-stage version labels to describe maturity. These are product-development labels, not claims about methodological completeness.

v0

Concept scaffold.

  • Skills exist.
  • Scope is clear.
  • Guardrails are explicit.
  • Methods are referenced.
  • No real execution path is required.

Meaning: the project idea is worth testing.

v0.1

Method scaffold.

  • README and positioning are clear.
  • Skills are split by task.
  • References and templates support progressive disclosure.
  • Companion boundary with EmpiriForge is clear.

Meaning: the project can guide research planning and audit conversations.

v0.2

Minimal executable spine.

  • Machine-readable coding schema.
  • Example coding sheet.
  • Minimal metafor script.
  • Machine-readable screening log schema.
  • Golden examples.

Meaning: the project can demonstrate a small protocol-to-data-to-output loop.

v0.3

Validation layer.

  • Coding-sheet validator.
  • Screening-log validator.
  • R environment notes.
  • GitHub Actions checks.
  • Effect-size helper functions.
  • IPD/mega-analysis templates for harmonized data integration.

Meaning: the project can reject some bad inputs before analysis.

v0.4

Reporting utilities.

  • PRISMA-style flow generator.
  • More golden examples.
  • Clearer output/report templates.
  • Better executable documentation.

Meaning: the project can produce auditable review artifacts, not just advice.

v0.5

Usability release.

  • Installation instructions tested on common platforms.
  • Example workflows run end to end.
  • Basic release notes.
  • Stable folder structure.
  • More realistic case examples.

Meaning: the project is ready for broader trial use.

v1.0

Stable public version.

  • Core skill interfaces are stable.
  • Deterministic scripts have tests.
  • Example workflows are reproducible.
  • Documentation explains boundaries clearly.
  • The project is safe to recommend as a serious evidence-synthesis workflow toolkit.

Meaning: the project is no longer only an experimental scaffold.

Current Direction

EvidenceForge is currently best described as v0.4 — the v0.3 validation layer and v0.4 reporting utilities are now complete:

  • it now has schemas, validators, golden examples, a minimal R script, effect-size helpers, and a PRISMA-style flow generator;
  • it also has an IPD/mega-analysis reference and audit templates inspired by small-sample multi-site synthesis workflows;
  • it is expanding the environmental branch with PLS/VIP environmental indicator audits, ecosystem-service relationship threshold ML audits, air-quality food-security co-benefit audits, biodiversity-stability climate-stress audits, soil-fauna carbon meta-analysis templates, urban heat DML, spatial causal-inference guardrails, agricultural irrigation optimization templates, and cryosphere ground-ice map-product audits;
  • it now has high-value article reproducibility audits that prioritize analysis-form fit, data-table structure, rma.mv, random forest/BRT, PLS-PM/SEM-family models, peer-review lessons, and reusable skill extraction;
  • it has a local meta-literature inventory with several strong case candidates, including Nature Communications potassium fertilization, ant-mediated soil carbon, agricultural diversification second-order meta-analysis, a Nature Medicine mega-analysis, and a Communications Psychology Bayesian meta-analysis;
  • it still needs richer effect-size conversion, robust/multilevel model templates, PRISMA/report export polish, and real case studies before v1.0.

To reach v0.5, at least two or three representative case studies should be runnable or auditable end to end:

  • one first-order treatment-control meta-analysis with open CSV/R code;
  • one ecological meta-analysis with multilevel dependence, ML ranking, and path modeling;
  • one second-order meta-analysis with overlap, quality, duration, and trade-off modeling.

The current recommended v0.5 case-study package is:

  • potassium fertilization and yield/SOC, for a first-order treatment-control nutrient meta-analysis;
  • riparian buffers and biodiversity, for an ecological meta-analysis where processed effect-size data and R code are public even when raw species-by-site data are protected;
  • agricultural diversification, for second-order meta-analysis and trade-off modeling.

See docs/v0.5-reproducible-case-study-candidates.md.