Deborah.Miriam

module Miriam

Deborah.Miriam — Multi-ensemble reweighting & interpolation for lattice-QCD cumulants.

Deborah.Miriam aggregates single-ensemble outputs, performs multi-ensemble reweighting and interpolation along $\kappa$-trajectories, and emits jackknife/bootstrap summaries for traces, moments, and higher-order cumulants ($\Sigma$, $\chi$, $S$, $K$, $B$). It is the multi-ensemble layer that sits downstream of Deborah.Esther/Deborah.DeborahCore and prepares analysis-ready text files and JLD2 snapshots for reporting and figure generation.

Scope & Responsibilities

• Configuration: parse TOML into strongly-typed structs ([data], [input_meta], [solver], [bootstrap], [jackknife], [trajectory], and [abbreviation]).
• Reweighting curves & transition points: scan $\kappa$, build bootstrap distributions for OG/P1/P2 estimators, and locate transition $\kappa$ via susceptibility/skewness/kurtosis probes; write per-$\kappa$ tables and final $\kappa_t$ with errors.
• Resampling outputs:
  • Bootstrap: traces ($\text{Tr}\,M^{-n} \; (n=1,2,3,4)$), moments ($Q_n \; (n=1,2,3,4)$), and cumulants → three files (OG/P1/P2) with $\mu \pm \sigma$ per $\kappa$.
  • Jackknife: traces, moments, and cumulants → $\mu \pm \sigma$ tables per $\kappa$.
• Path-ing & names: stable, abbreviation-aware filenames/dirs driven by the config.

Key Components

• Deborah.Miriam.TOMLConfigMiriam — types & TOMLConfigMiriam.parse_full_config_Miriam for [data], [input_meta], [solver], [bootstrap], [jackknife], [trajectory], and [abbreviation].
• ReweightingCurveBundle.reweighting_curve_bundle! — $\kappa$-scan with bootstrap OG/P1/P2, transition-point interpolation, and file emission.
• Deborah.Miriam.WriteBSOutput — write_bs_traces, write_bs_moments, write_bs_cumulants (bootstrap mean & error per $\kappa$ for OG/P1/P2).
• Deborah.Miriam.WriteJKOutput — write_jk_traces, write_jk_moments, write_jk_cumulants (jackknife mean & error per $\kappa$).

Public API (typical entry points)

• MiriamRunner.run_Miriam (orchestrated run: load → resample → reweight/interpolate → write tables).

Minimal Usage (REPL)

julia> using Deborah
julia> run_Miriam("config_Miriam.toml")

Notes

• Estimators: OG (original), P1 (bias-corrected), P2 (weighted LB and P1).
• Bootstrap: supports nonoverlapping / moving / circular block schemes; reuse of bootstrap indices ensures fair OG/P1/P2 comparison.
• Jackknife: delete-1 with binning; volume factor $V = N_S^3 \times N_T$ used where needed.
• All column indices are $1$-based; abbreviation maps, when enabled, shorten tokens in names and paths consistently.

See Also

• Deborah.DeborahCore: trace estimation
• Deborah.Esther: single ensemble cumulants
• Deborah.DeborahEstherMiriam: pipeline bridge
• Deborah.MiriamDocument: report layer