Class Definitions

The Python port preserves the MATLAB-facing class names as canonical imports or compatibility adapters. Each class below lists its key public methods grouped by category.

Signal and Covariate Primitives

SignalObj (nstat.SignalObj)

Primary notebook: ../notebooks/SignalObjExamples.ipynb

Construction and metadata: copySignal, setName, setXlabel, setYLabel, setUnits, setXUnits, setYUnits, setSampleRate, setDataLabels, setPlotProps, with_metadata

Accessors: getTime, getData, getOriginalData, getOrigDataSig, getPlotProps, getValueAt, getSubSignalFromInd, getSubSignalFromNames, getSubSignal, findNearestTimeIndex, findNearestTimeIndices, dataToMatrix, dataToStructure, toStructure, signalFromStruct

Masking: setDataMask, setMaskByInd, setMaskByLabels, setMask, resetMask, restoreToOriginal, findIndFromDataMask, isMaskSet

Time windowing: setMinTime, setMaxTime, getSigInTimeWindow, makeCompatible

Math and transforms: abs, log, power, sqrt, median, mode, mean, std, max, min, derivative, derivativeAt, integral, resample, resampleMe, filter, filtfilt, merge

Shift and alignment: shift, shiftMe, alignTime

Correlation: autocorrelation, crosscorrelation, xcorr, xcov

Spectral analysis: periodogram, MTMspectrum, spectrogram

Peak-finding: findPeaks, findMaxima, findMinima, findGlobalPeak

Plotting: plot

Covariate (nstat.Covariate)

Primary notebook: ../notebooks/CovariateExamples.ipynb

Inherits from SignalObj. Adds confidence interval support: setConfInterval, mu, sigma

nstat.CovariateCollection is the Pythonic alias for CovColl (below).

ConfidenceInterval (nstat.ConfidenceInterval)

Primary notebook: ../notebooks/ConfidenceIntervalOverview.ipynb

CovColl (nstat.CovColl)

Primary notebook: ../notebooks/CovCollExamples.ipynb

Collection management: add, addCovariate, addCovCollection, addToColl, removeCovariate, copy, get, getCov, getCovIndFromName, getCovIndicesFromNames, isCovPresent

Time and sample rate: findMinTime, findMaxTime, findMaxSampleRate, setMinTime, setMaxTime, restrictToTimeWindow, setSampleRate, resample, enforceSampleRate

Masking: resetMask, getCovDataMask, isCovMaskSet, flattenCovMask, getSelectorFromMasks, generateSelectorCell, setMasksFromSelector, setMask, nActCovar, maskAwayCov, maskAwayOnlyCov, maskAwayAllExcept

Shift and restore: setCovShift, resetCovShift, restoreToOriginal

Data export: getAllCovLabels, getCovLabelsFromMask, matrixWithTime, dataToMatrix, dataToStructure, toStructure, fromStructure

Spiking Data Structures

nspikeTrain (nstat.nspikeTrain)

Primary notebook: ../notebooks/nSpikeTrainExamples.ipynb

nstat.SpikeTrain is the Pythonic alias for nspikeTrain.

nstColl (nstat.nstColl)

Primary notebook: ../notebooks/nstCollExamples.ipynb

nstat.SpikeTrainCollection is the Pythonic alias for nstColl.

Collection management: addSingleSpikeToColl, addToColl, merge, length, get_nst, getNST, getNSTnames, getUniqueNSTnames, toSpikeTrain

Time operations: shiftTime, setMinTime, setMaxTime

Data export: dataToMatrix, resample

PSTH: psth, psthGLM, estimateVarianceAcrossTrials, psthBars

SSGLM (state-space GLM): ssglm, ssglmFB

Basis generation: generateUnitImpulseBasis

Plotting: plot

History (nstat.History)

nstat.HistoryBasis is a companion dataclass exposing the basis-function parameters used to build self-history kernels (window times, basis family, raised-cosine vs unit-impulse choices).

Primary notebook: ../notebooks/HistoryExamples.ipynb

Events (nstat.Events)

Primary notebook: ../notebooks/EventsExamples.ipynb

Experiment and Configuration Objects

Trial (nstat.Trial)

Primary notebook: ../notebooks/TrialExamples.ipynb

Partitioning: getTrialPartition, setTrialPartition, setTrialTimesFor, updateTimePartitions

Time and sample rate: setMinTime, setMaxTime, setSampleRate, resample, makeConsistentSampleRate, makeConsistentTime, isSampleRateConsistent, findMaxSampleRate, findMinTime, findMaxTime

Covariate and neuron masks: setCovMask, resetCovMask, setNeuronMask, resetNeuronMask, setNeighbors, setHistory, resetHistory, setEnsCovHist, setEnsCovMask, resetEnsCovMask, isNeuronMaskSet, isCovMaskSet, isMaskSet, isHistSet, isEnsCovHistSet

Data access: addCov, removeCov, getSpikeVector, get_covariate_matrix, getDesignMatrix, getHistForNeurons, getHistMatrices, getEnsembleNeuronCovariates, getEnsCovMatrix, getNeuronIndFromMask, getNumUniqueNeurons, getNeuronNames, getUniqueNeuronNames, getNeuronIndFromName, getAllCovLabels, getCovLabelsFromMask, getHistLabels, getEnsCovLabels, getLabelsFromMask, flattenCovMask, flattenMask

Utilities: shiftCovariates, resampleEnsColl, restoreToOriginal, getAllLabels, plot

TrialConfig (nstat.TrialConfig)

Primary notebook: ../notebooks/TrialConfigExamples.ipynb

ConfigColl (nstat.ConfigColl)

Primary notebook: ../notebooks/ConfigCollExamples.ipynb

nstat.ConfigCollection is the Pythonic alias for ConfigColl.

Modeling and Inference

CIF (nstat.CIF)

Primary notebook: ../notebooks/PPSimExample.ipynb

Conditional intensity function (CIF) primitives and point-process simulation via thinning.

nstat.CIFModel is a lightweight dataclass capturing the parameters of a fitted CIF — coefficients, link function, history kernel. Used as a return type by some Analysis paths.

LinearCIF (nstat.LinearCIF)

Closed-form, sympy-free CIF for the two canonical link cases (Poisson log-link, binomial logit-link). Ported from MATLAB LinearCIF.m (added upstream in v1.4.0). Drop-in compatible with CIF for the 5 eval methods used by DecodingAlgorithms.PPDecode_update: evalLambdaDelta, evalGradient, evalGradientLog, evalJacobian, evalJacobianLog.

Analysis (nstat.Analysis)

Primary notebook: ../notebooks/AnalysisExamples.ipynb

Core fitting: GLMFit, run_analysis_for_neuron, run_analysis_for_all_neurons, RunAnalysisForNeuron, RunAnalysisForAllNeurons

PSTH: psth

Diagnostics: computeKSStats, computeInvGausTrans, computeFitResidual, KSPlot, plotFitResidual, plotInvGausTrans, plotSeqCorr, plotCoeffs

History and model selection: computeHistLag, computeHistLagForAll, compHistEnsCoeff, compHistEnsCoeffForAll

Network and Granger causality: computeGrangerCausalityMatrix, computeNeighbors, spikeTrigAvg

FitResult (nstat.FitResult)

Primary notebook: ../notebooks/FitResultExamples.ipynb

Coefficient access: getCoeffs, getHistCoeffs, getCoeffIndex, getHistIndex, getParam, getCoeffsWithLabels, computePlotParams, getPlotParams

Lambda (conditional intensity) access: lambdaSignal, lambda_obj, lambda_model, lambda_result, lambdaObj, lambdaCov, lambda_sig, lambda_data, lambda_values, lambda_time, lambda_rate, evalLambda

Diagnostics and statistics: computeKSStats, computeInvGausTrans, computeFitResidual

Plotting: plotResults, KSPlot, plotResidual, plotInvGausTrans, plotSeqCorr, plotCoeffs, plotCoeffsWithoutHistory, plotHistCoeffs, plotValidation

Serialization: toStructure, fromStructure, CellArrayToStructure, mergeResults

FitResSummary (nstat.FitResSummary)

Primary notebook: ../notebooks/FitResSummaryExamples.ipynb

Alias of FitSummary. Aggregates multiple FitResult objects.

Information criterion: getDiffAIC, getDiffBIC, getDifflogLL

Coefficient extraction: getCoeffs, getHistCoeffs, getSigCoeffs, binCoeffs, setCoeffRange, mapCovLabelsToUniqueLabels

Plotting: plotIC, plotAIC, plotBIC, plotlogLL, plotResidualSummary, plotSummary, boxPlot, plotAllCoeffs, plot3dCoeffSummary, plot2dCoeffSummary, plotKSSummary

PopulationTimeRescaleResult (nstat.PopulationTimeRescaleResult)

Frozen dataclass returned by population_time_rescale(...) — the multivariate (marked) point-process time-rescaling goodness-of-fit of Tao, Weber, Arai & Eden (2018). Unlike FitResult.computeKSStats (per-neuron, univariate), it scores a population jointly and catches inter-neuron coupling misfit (e.g. synchronous neurons modeled as independent). Python-only extension — no MATLAB matlab_gold counterpart.

Fields:

  • ground_uniforms — rescaled ground-process inter-event values on [0, 1].

  • ground_ks_stat, ground_ks_pvalue — KS of the pooled (ground) process vs. Uniform(0, 1); detects aggregate-temporal / dependency misfit.

  • mark_chi2_stat, mark_chi2_dof, mark_chi2_pvalue — Pearson χ² for uniform fill of the marked region R = {(τ, m): 0 ≤ τ ≤ b(m)}; detects relative-allocation (and, with n_tau_bins > 1, within-neuron timing) misfit.

  • expected_counts, observed_counts — per-neuron ∫λ vs. observed spikes.

DecodingAlgorithms (nstat.DecodingAlgorithms)

Primary notebook: ../notebooks/DecodingExample.ipynb

Point-process decode filters: PPDecodeFilterLinear, PPDecodeFilter, PPHybridFilterLinear, ComputeStimulusCIs, computeSpikeRateCIs

Kalman and unscented Kalman filters: kalman_filter, PP_fixedIntervalSmoother, ukf

Helper methods: PPDecode_predict, PPDecode_updateLinear

State-space GLM (SSGLM) — EM algorithm: PPSS_EStep, PPSS_MStep, PPSS_EM, PPSS_EMFB

SSGLM utilities: estimateInfoMat, prepareEMResults

DecoderSuite (nstat.DecoderSuite)

Pythonic wrapper that exposes the most common DecodingAlgorithms static methods through a friendlier object-oriented API. No MATLAB counterpart — this is a Python convenience layer over the parity-ported algorithm namespace.

PoissonGLMResult (nstat.PoissonGLMResult)

Dataclass returned by :func:nstat.fit_poisson_glm. Captures the fitted coefficients, intercept, deviance, iteration count, convergence status, log-likelihood, and standard errors of a standalone Poisson GLM fit (separate from Analysis.GLMFit, which is wired through the Trial machinery).

Simulation Objects

PointProcessSimulation (nstat.PointProcessSimulation)

Result dataclass for :func:nstat.simulate_point_process. Holds the simulated spike train, the underlying conditional intensity trace, and the seed / RNG state for reproducibility.

NetworkSimulationResult (nstat.NetworkSimulationResult)

Result dataclass for :func:nstat.simulate_two_neuron_network. Captures both neurons’ spike trains and the cross-coupling history that drove the simulation.

Spatial / spatiotemporal point processes

Python-only result types from nstat.extras.spatial. No MATLAB counterpart, no parity/manifest.yml entry; see docs/extras/spatial_point_processes.md for the full API and recipes.

BSplineBasis2D (nstat.extras.spatial.basis.BSplineBasis2D)

Frozen dataclass capturing a tensor-product B-spline log-rate basis on a rectangular grid. Constructed via BSplineBasis2D.from_grid(...); fed as the basis argument to lgcp_fit_glm.

Fields:

  • grid_x, grid_y (ndarray) — evaluation grids in x and y.

  • n_knots_x, n_knots_y (int) — number of basis functions in each axis (so the basis dimension is K = n_knots_x * n_knots_y).

  • degree (int) — B-spline degree (default 3, cubic).

  • clamped (bool) — clamped (open uniform) knot vector when True (partition-of-unity rows); periodic otherwise.

Methods:

  • from_grid(grid_x, grid_y, n_knots, degree=3, clamped=True) → BSplineBasis2D — class constructor.

  • design_matrix() → ndarray shape (Nx*Ny, K) — cached tensor-product design matrix in indexing="ij" row layout.

  • gram() → ndarray shape (K, K) — P-spline second-difference penalty Dx.T Dx ⊗ Iy + Ix ⊗ Dy.T Dy, symmetric PSD by construction (Eilers-Marx 1996).

  • coefficient_coords() → ndarray shape (K, 2) — Greville-abscissa anchor points of the coefficients in ij flattening; feed to MaternPrior to evaluate the GP prior on the coefficient vector.

References: de Boor C (1978), A Practical Guide to Splines, Springer; Eilers PHC, Marx BD (1996), Flexible Smoothing with B-splines and Penalties, Statistical Science 11(2):89-121.

MaternPrior (nstat.extras.spatial.MaternPrior)

Frozen dataclass holding a Matern Gaussian-process prior evaluated at a fixed coordinate set (typically BSplineBasis2D.coefficient_coords()). Caches K, its Cholesky factor, K_inv, and log_det per coords array.

Fields:

  • nu (float) — Matern smoothness parameter; supported nu ∈ {0.5, 1.5, 2.5}.

  • length_scale (float) — isotropic length scale.

  • marginal_var (float) — marginal variance of the GP (default 1.0).

  • jitter (float) — diagonal Cholesky jitter (default 1e-6); bumped 10x and retried once on a Cholesky failure.

Methods:

  • K(coords) → ndarray — covariance matrix at coords.

  • cholesky(coords) → ndarray — lower Cholesky factor of K(coords).

  • K_inv(coords) → ndarray — inverse covariance matrix.

  • log_det(coords) → float — log |K(coords)|.

Reference: Rasmussen CE, Williams CKI (2006), Gaussian Processes for Machine Learning, Algorithm 3.1.

WaveAnalysisResult (nstat.extras.spatial.WaveAnalysisResult)

Frozen dataclass returned by detect_wave_peaks(...). One entry per accepted spectral peak of a Hawkes triggering matrix’s Bartlett (frequency × wave-vector) spectrum.

Fields:

  • freq (ndarray) — temporal frequency of each peak (Hz).

  • kx, ky (ndarray) — wave-vector components at each peak.

  • power (ndarray) — spectral power at each peak.

  • speed (ndarray) — phase speed 2*pi*freq / |k| in position-unit / s.

  • direction (ndarray) — atan2(ky, kx) in radians.

Reference: Bacry E, Mastromatteo I, Muzy J-F (2015), Hawkes processes in finance, Market Microstructure and Liquidity 1(1):1550005; Daley DJ, Vere-Jones D (2003), An Introduction to the Theory of Point Processes, Vol. I, §8.4.

See Examples for the full help-style index and API Reference for the module layout.