Extras — opt-in bridges to the Python neuro ecosystem

The nstat.extras namespace ships Python-only features that have no counterpart in upstream MATLAB nSTAT. Each subpackage depends on an optional library declared in pyproject.toml — install via pip install nstat-toolbox[<group>] (e.g. [neo], [pynapple], [nwb], [metrics], [nemos], [test-parity], or [all-extras] for the union).

For the design rationale and stability contract, see nstat.extras and the integration_opportunities audit.

Narrative usage guides

Per-bridge documentation with install commands, API tables, recipes, gotchas, and links to the runnable demos under examples/extras/.

• nstat.extras.interop.neo — Neo bridge
• nstat.extras.interop.pynapple — pynapple bridge
• nstat.extras.interop.nwb — NWB:N reader
• nstat.extras.validation.nemos_bridge — NeMoS GLM cross-validation
• nstat.extras.validation.pykalman_bridge — pykalman Kalman cross-validation
• nstat.extras.validation.statsmodels_bridge — statsmodels GLM cross-validation
• nstat.extras.metrics.spike_distances — PySpike spike-train distances
• nstat.extras.em.dynamax_bridge — EM-trained state-space models via Dynamax
• nstat.extras.decoding.clusterless_bridge — Clusterless point-process decoding
• nstat.extras.decoding.place_field_decoder — Place-cell encoding + 2-D PPAF decoding
• nstat.extras.spatial — Spatial & spatiotemporal point processes
• nstat.extras.latents.gpfa_bridge
• nstat.extras.matlab_rng — MATLAB-aligned Mersenne Twister RNG
• nstat.extras.continuous_cif — continuous-time CIF simulator

Interop — data-model bridges

interop.neo	Neo ↔ nstat converters (Tier-B interop, scope: spike trains).
interop.pynapple	pynapple ↔ nstat converters (Tier-B interop).
interop.nwb	NWB (Neurodata Without Borders) → nstat reader (Tier-B interop).

Validation — cross-validation oracles

validation.nemos_bridge	Cross-validate nstat.fit_poisson_glm() against NeMoS.
validation.pykalman_bridge	Cross-validate nstat.DecodingAlgorithms.kalman_filter() against pykalman.
validation.statsmodels_bridge	Cross-validate nstat.fit_poisson_glm() against statsmodels.

Metrics — modern spike-train distance metrics

metrics.spike_distances

Spike-train distance metrics — ISI, SPIKE, SPIKE-synchronization.

EM — state-space models trained by expectation-maximization

em.dynamax_bridge

EM-trained linear-Gaussian state-space models via Dynamax.

Decoding — Bayesian point-process decoders

decoding.clusterless_bridge	Clusterless / trajectory-classification decoding via replay_trajectory_classification.
decoding.place_field_decoder	High-level place-field encoding + decoding wrapper.

Spatial — spatial & spatiotemporal point processes

spatial.lgcp	Log-Gaussian Cox process rate maps by the Laplace approximation.
spatial.spatial_gof	Inhomogeneous second-order spatial goodness-of-fit (pure NumPy/SciPy).
spatial.marked_gof	Marked / discrete-time-rescaling goodness-of-fit (pure NumPy/SciPy).
spatial.hawkes_bridge	Multivariate Hawkes process bridge via tick (optional).
spatial.dpp_bridge	Determinantal point process (DPP) sampling — DPPy bridge + NumPy fallback.

Latents — latent-variable / dimensionality-reduction bridges

latents.gpfa_bridge

Gaussian-Process Factor Analysis bridge via Elephant.