timescales.fit.ARPSD

class timescales.fit.ARPSD(order, fs, bounds=None, ar_bounds=None, guess=None, maxfev=100, loss_fn='linear', scaling='log', f_scale=None, curve_fit_kwargs=None)

Fits AR(p) model to PSD.

__init__(order, fs, bounds=None, ar_bounds=None, guess=None, maxfev=100, loss_fn='linear', scaling='log', f_scale=None, curve_fit_kwargs=None)

Intialize object.

Parameters:

orderint

Autoregressive order.

fsfloat

Sampling rate, in Hertz.

bounds2d tuple or list, optional, default: None

Bounds on the AR weights as (lower, uper). Defaults to (-0.9999, 0.9999). In some cases, (0, 0.9999) may be more appropriate.

ar_boundstuple of (float, float):

Sets bounds across all AR weights.

guesslist, optional, default: None

Inital AR weights. Defaults to zeros.

maxfevint, optional, default: None

Max number of optimization iterations.

loss_fnstr, optional, default: ‘linear’

Name of loss function supported by curve_fit.

scalingstr, {“log”, “linear”}

Scale of power to fit.

f_scalefloat, optional, default: None

Robust regression. Determines inliers/outliers. Between [0, 1].

curve_fit_kwargsdict, optional, default: None

Additonal kwargs to pass to curve_fit.

Methods

__init__(order, fs[, bounds, ar_bounds, ...])	Intialize object.
fit(freqs, powers)	Fit PSD.
plot([ax])	Plot model fit.
simulate(n_seconds, fs[, init, error, index])	Simulate a signal based on learned parameters.

Attributes

is_stationary

Determines if the learned coefficients give a stationary process.

fit(freqs, powers)

Fit PSD.

Parameters:

freqs1d array

Frequencies.

powers1d or 2d array

Power.

property is_stationary

Determines if the learned coefficients give a stationary process.

plot(ax=None)

Plot model fit.

simulate(n_seconds, fs, init=None, error=None, index=None)

Simulate a signal based on learned parameters.