bootstrap_ar2() and fit_ar2() no longer
restore .Random.seed in the global environment after a
seeded bootstrap. Supplying seed now simply calls
set.seed(seed); leave it NULL to draw from the
caller’s stream.CORE_CLOCK_GENES and
KNOWN_TARGET_GENES.fit_ar2() no longer caps |lambda| at 2. An
unstable fit now reports the spectral radius of its fitted coefficients,
matching par2-circadian 1.2.1. Stable fits (|lambda| <
1) are unchanged.First release: the R implementation of the AR(2) eigenvalue persistence method, ported from the Python package par2-circadian 1.2.0.
fit_ar2() for one series and
fit_ar2_batch() for a genes-by-timepoints matrix, returning
|lambda|, the AR(2) coefficients, R^2, root type, half-life
and eigenperiod.bootstrap_ar2() residual-bootstrap confidence
intervals, reachable through fit_ar2(n_bootstrap =), for
the short series circadian designs produce.discover_hierarchy(), gearbox_gap() and
classify_gene_layer() for the Clock > Target >
Background stratification; classify_dynamics() for
root-type-aware labels.half_life() and eigenperiod() convert to
real time units.