Methods

Flagship methods

Method	Input	Maturity	Backend story	Why start here
SSA	1D	flagship	native C++ plus Python fallback	strong first choice for interpretable single-series decomposition
STD	1D and channelwise 2D	flagship	native C++ plus Python fallback	simple seasonal-trend separation with stable defaults
STDR	1D and channelwise 2D	flagship	native C++ plus Python fallback	robust seasonal-trend decomposition with shared seasonal-shape estimation
MSSA	2D	flagship	Python implementation	joint multichannel decomposition when channels share structure

Retained wrappers and specialist paths

Method	Input	Maturity	Dependency tier	Notes
STL	univariate	stable	core-upstream	good baseline when the primary period is known
MSTL	univariate	stable	core-upstream	multiple seasonal periods
EMD	univariate	stable	core	adaptive decomposition
CEEMDAN	univariate	stable	core	ensemble adaptive decomposition
VMD	univariate	stable	core	variational mode decomposition
WAVELET	univariate	stable	core-upstream	multi-scale decomposition
MA_BASELINE	univariate	stable	core	lightweight sanity check
MVMD	multivariate	optional-backend	optional-backend	install with the multivar extra
MEMD	multivariate	optional-backend	optional-backend	install with the multivar extra
GABOR_CLUSTER	univariate	experimental	core	use after you already trust a baseline

Moved out of the main package

Benchmark-derived methods are not part of detime. Companion benchmark work now lives in systems-mechanobiology/de-time-bench.

How to read this surface

• Start with SSA, STD, STDR, or MSSA unless a specialist method is clearly required.
• Treat upstream wrappers as integration convenience, not as evidence that De-Time replaces the upstream package.
• Treat optional multivariate backends as opt-in extras.
• Use detime recommend when you want a machine-readable shortlist instead of choosing manually.