Results

The main output of jacscanomaly.Finder.run() is jacscanomaly.AnomalyResult.

Summary methods

AnomalyResult provides CLI-friendly and notebook-friendly summaries:

result.print_summary()
text = result.summary_text()
row = result.summary_dict()
table = result.summary_table()

summary_table returns a one-row pandas.DataFrame when pandas is installed. Otherwise it returns list[dict].

Important attributes

time, flux, ferr

Input arrays stored as NumPy arrays.

fit

The single-lens fit result.

model_flux

Baseline single-lens model flux.

residual

flux - model_flux.

chi2_dof

Reduced chi-square of the baseline fit.

seasons

List of jacscanomaly.SeasonSummary objects.

clusters_all

Flattened extracted clusters. Rows are [t0, teff, dchi2].

grid_metrics_all

Flattened per-grid diagnostics. Columns are:

[t0, teff, dchi2, n_window, n_contrib, n_eff, peak_frac, rho1, longest_run]

best

The best jacscanomaly.BestCandidate, or None if no candidate exists.

Best candidate

if result.best is not None:
    best = result.best
    print(best.t0)
    print(best.teff)
    print(best.dchi2)
    print(best.score)

    quality = best.quality
    print(quality.n_eff)
    print(quality.peak_frac)

Candidate quality fields

n_window: Number of points in the local anomaly window.
n_contrib: Number of points above the per-point contribution threshold.
n_eff: Effective number of contributing points. This is often the most useful diagnostic for rejecting one-point artifacts.
peak_frac: Strongest-point contribution divided by total positive contribution.
rho1: Lag-1 autocorrelation of per-point improvements.
longest_run: Longest run of above-threshold contributing points.

Working with all candidates

The finder stores non-overlapping cluster representatives in clusters_all:

t0 = result.clusters_all[:, 0]
teff = result.clusters_all[:, 1]
dchi2 = result.clusters_all[:, 2]

The raw grid diagnostics are in grid_metrics_all:

metrics = result.grid_metrics_all
n_eff = metrics[:, 5]
supported = metrics[n_eff > 2.0]

This is useful when building survey-level candidate tables or applying custom post-processing criteria.