Long Memory · Network-Adaptive Efficiency (NAMH)
namh_hurst
NAMH Hurst Panel (published namh). Rolling-window DFA-ℓ Hurst panel (Peng et al. 1994) via the PUBLISHED namh package (Bhandari & Sahu 2026) — the paper's own estimate_hurst_panel, not a reimplementation. Supplies the NAMH node-weight φ(H)=1−2|H−0.5| (local efficiency). Defaults reproduce the canonical paper-v3 panel on g20_24 (window=252, step=252 non-overlapping, DFA-1): bit-exact to 01_hurst_panel.csv (max|Δ|≈5e-9). Per-series summary by default; runs all 24 series if none given.
Identity
version 1.0.0 · capability long-memory · min_obs 252 · runner r (namh_hurst.R)
Primitives
P1
Composes with
not yet a catalog field — operator-composition metadata lands with Pathway F; nothing is invented here.
Parameters
| name | type | req | values / range |
|---|
| series | series (n=1–24) | optional | |
| window | int | optional | |
| step | int | optional | |
| order | int | optional | |
| s_min | int | optional | |
| n_scales | int | optional | |
Returns
method,dataset,config,n_series,n_windows,per_series,source
Paper
none — honest: a substrate method; no single companion paper claims it.
Changelog
[object Object]
Run it
curl -s -X POST https://shssm-compute-b7ui3oxaqq-el.a.run.app/api/compute/run \
-H "Content-Type: application/json" \
-d '{"method":"namh_hurst","params":{"series":["India","USA"]}}'