ssfortran.Trend#

class ssfortran.Trend(level_cov='diagonal', slope_cov='diagonal', at_observations=False)#

Local linear trend (DK §3.2.1).

\(\mu_{t+1} = \mu_t + \nu_t + \xi_t\), \(\nu_{t+1} = \nu_t + \zeta_t\).

Parameters:
level_cov, slope_cov{“diagonal”, “full”, None}

Forms of \(Var(\xi_t)\) and \(Var(\zeta_t)\). level_cov=None gives the smooth trend (integrated random walk); both None a deterministic linear trend.

at_observationsbool

Act on the observations rather than the signals.