posterior_draws

SuperLSS.posterior_draws(

n_draws: int = 1000,

*,

covariance: Literal['fixed', 'corrected'] = 'fixed',

seed: int = 42,

) → PosteriorDraws

Return coefficient draws from the Bayesian posterior of the fit.

The draws are N(beta_hat, V) in the fit’s own global coordinates, which is the posterior of Marra and Wood (2012) with the smoothing parameters held at their estimates; covariance="corrected" asks instead for the smoothing-uncertainty correction of Wood, Pya and Säfken (2016), using either trusted published curvature or one authenticated replay from a stationary fit’s retained rows. Compact fits without published curvature and untrusted fits refuse. One draw set can be reused across quantities.