Deborah.Esther

module Esther

Deborah.Esther — Higher-order cumulant analysis for lattice-QCD trace data.

Deborah.Esther implements the statistical pipeline to extract chiral-condensate observables from trace moments $\text{Tr}\,M^{-n} \; (n=1,2,3,4)$: loading and rescaling traces, computing $Q$-moments and derived cumulants, running bootstrap/jackknife, and emitting formatted summaries and files ready for downstream reporting.

Scope & Responsibilities

• Configuration & paths

  • Deborah.Esther.TOMLConfigEsther: parse a single TOML into strongly-typed structs (TraceDataConfig, InputMetaConfig, BootstrapConfig, JackknifeConfig, FullConfigEsther), including abbreviation maps.
  • Deborah.Esther.PathConfigBuilderEsther: construct reproducible output layouts and names.
  • Deborah.Esther.DatasetPartitionerEsther: prepare (labels $\times$ trains) partitions and indices.

• Trace I/O & normalization

  • Deborah.Esther.TraceDataLoader: load Y_* / YP_* series for $\text{Tr}\,M^{-n} \; (n=1,2,3,4)$ into label-keyed dictionaries (Y_info, Y_tr, Y_bc, Y_ul, Y_lb, YP_tr, YP_bc, YP_ul).
  • Deborah.Esther.TraceRescaler: apply $V$/$\kappa$ rescaling for each power $p$: $[\mathrm{Tr}\,M^{-p}]_{\mathrm{rescaled}} = 12V(2\kappa)^p [\mathrm{Tr}\,M^{-p}]_{\mathrm{in}}$.

• Moment & cumulant computation

  • Deborah.Esther.SingleQMoment: compute per-configuration $Q_n \; (n=1,2,3,4)$ from $\text{Tr}\,M^{-n} \; (n=1,2,3,4)$ and $N_{\text{f}}$.
  • Deborah.Esther.QMomentCalculator: vectorized $Q$-moment builders over resamples.
  • Deborah.Esther.SingleCumulant: per-resample chiral condensate / susceptibility / skewness / kurtosis from $Q_n \; (n=1,2,3,4)$ with the standard formulas.
  • Deborah.Esther.BootstrapDerivedCalculator: derive observables from bootstrap-averaged $Q$-data.

• Resampling & error analysis

  • Deborah.Esther.JackknifeRunner: delete-$1$ jackknife with binning support (bin size from gauge configuration).
  • Bootstrap methods: nonoverlapping / moving / circular (config-driven).

• Reporting & output

  • Deborah.Esther.SummaryWriterEsther: write summary_Esther_<overall_name>.dat for models ($\text{Tr}\,M^{-n} \; (n=1,2,3,4)$), $Q_n \; (n=1,2,3,4)$, and derived observables with (mean, std) per label/tag.
  • Deborah.Esther.ResultPrinterEsther: print console tables grouping jackknife/ bootstrap by tags (e.g., Y:JK, Y:BS, Y_P1, Y_P2, etc.).

• Execution

  • Deborah.Esther.EstherRunner: end-to-end runner (config → load/scale traces → $Q$-moments → cumulants → resampling → summaries).

Public API (typical entry point)

• EstherRunner.run_Esther

Minimal Usage (REPL)

julia> using Deborah
julia> run_Esther("config_Esther.toml")

Notes

• Column indices in configs are $1$-based and propagated as-is.
• Output directory scheme is derived from analysis_header and ensemble to keep results reproducible across runs.
• Abbreviation maps (when enabled) shorten feature/target tokens consistently across file and folder names.

See Also

• Deborah.DeborahCore (trace estimation with ML/bias correction),
• Deborah.DeborahEsther (bridge runner),
• Deborah.EstherDocument (report layer),
• Deborah.EstherThreads (threaded batch runner).