20 Methodology
This page indexes where each part of the methodology already lives rather than restating it — the detail is maintained in research-platform/ and phd/, generated from or grounded in the platform itself.
20.1 Data sources
- Data Catalog — provenance, licensing, refresh schedules for every dataset the research draws on.
- The multiplex data substrate — whether the platform can source the proposal’s three network layers (factor exposure, supply-chain, institutional ownership).
- PhD proposal §“Data & reproducibility” — public and commercial sources named for the research specifically (French data library, CRSP-like sources, SEC 13F, SEC 10-K parsing).
20.2 Statistical methods
- Factor models & benchmarks — factor loadings & returns, the CAPM decomposition, systematic vs. idiosyncratic return, benchmark selection.
- The factor layer as multiplex Layer 1 — the bipartite stock–factor layer as built, not hypothetical.
- Systemic-risk & crowding groundwork — prior network work (comomentum, crowding) the proposal builds on.
- Whitepaper §5 — the mathematics of trust — the statistical core (Information Coefficient, permutation testing, false-discovery correction, overfitting probability) that any candidate signal — including a network-based early-warning indicator — must clear.
20.3 Validation approach
- Validation & anti-overfitting — how network claims specifically are protected from self-deception, beyond the platform’s existing factor-validation gates.
- Reproducibility & experiment infrastructure — what makes an experiment auditable and repeatable here.
- Gap analysis & research roadmap — honest
BUILT/PARTIAL/PROPOSEDstatus per capability, so methodology claims don’t outrun what’s actually implemented.
See also: Research Questions for what each of these methods is being used to answer, and Experiments for the running record of applying them.