sidResidual¶

Python equivalent: sid.residual

Compute model residuals and perform diagnostic tests.

result = sidResidual(model, y, u)
result = sidResidual(model, y, u, 'MaxLag', M)
sidResidual(model, y, u, 'Plot', true)

Computes residuals from an estimated model and performs whiteness and independence tests to assess model quality.

Inputs¶

Name	Description
`model`	Result struct from any sid estimator (see sidResultTypes). Freq-domain (§1): requires .Response, .Frequency, .SampleTime State-space (§4/§5): requires .A, .B, .StateDim, .InputDim, .DataLength
`y`	(N x ny) measured output (or (N+1 x p x L) state data for COSMIC)
`u`	(N x nu) input, or [] for time-series models

Name	Description
`'MaxLag'`	Maximum lag for correlation tests (default: min(25, floor(N/5)))
`'Plot'`	Display diagnostic plot (default: true if nargout==0)

Name	Description
`result.Residual`	(N x ny) residual time series e(t)
`result.AutoCorr`	(M+1 x 1) normalised autocorrelation r_ee(tau)
`result.CrossCorr`	(2M+1 x 1) normalised cross-corr r_eu(tau), or []
`result.ConfidenceBound`	scalar, 99% bound 2.58/sqrt(N)
`result.WhitenessPass`	logical, true if autocorrelation test passes
`result.IndependencePass`	logical, true if cross-correlation test passes
`result.DataLength`	N

% Residual analysis for a frequency-domain model
G = sidFreqBT(y, u);
result = sidResidual(G, y, u);

% Visual diagnostic plot
sidResidual(G, y, u, 'Plot', true);

(Model residual analysis — not yet in SPEC.md)