geossm.stmodel.ModelParams

class geossm.stmodel.ModelParams(beta=None, s2e=None, f=None, A=None, ks=None, x0=None, Sigma0=None)

Bases: object

The full set of parameters estimated by fit().

Each field is a Param (wrapping a value, fixed/free status, and standard error); passing a plain array-like instead of a Param for any field is allowed and is converted to a free (non-fixed) Param automatically (see _ensure_param). Registered as a JAX pytree, so a ModelParams can be passed directly through jax.grad/jax.jit.

Parameters:

• beta (Param or array_like, optional) – Regression coefficients for the model’s exogenous covariates.

• s2e (Param or array_like, optional) – Measurement-error (observation noise) variance(s).

• f (Param or array_like, optional) – AR(1) coefficient(s) of the latent spatial factor(s)’ temporal evolution.

• A (Param or array_like, optional) – Loading matrix linking the latent spatial factor(s) to the observations.

• ks (Param or array_like, optional) – Matérn rescale/range parameter(s) of the latent covariance function(s) (one per factor).

• x0 (Param or array_like, optional) – Mean of the initial latent state, as in StateSpaceModel.

• Sigma0 (Param or array_like, optional) – Covariance of the initial latent state, as in StateSpaceModel.

Methods

as_dict	dict: Map each field name to its Param.value (fixed and free).
copy	Return a deep copy of the ModelParams.
free_params	dict: Map each non-fixed field name to its Param.value.
tree_flatten
tree_unflatten

Attributes

A
Sigma0
beta
f
ks
s2e
shape
size
x0

beta: Param | None = None

s2e: Param | None = None

f: Param | None = None

A: Param | None = None

ks: Param | None = None

x0: Param | None = None

Sigma0: Param | None = None

as_dict()

dict: Map each field name to its Param.value (fixed and free).

free_params()

dict: Map each non-fixed field name to its Param.value.

property size

property shape

tree_flatten()

classmethod tree_unflatten(aux_data, children)

copy()

Return a deep copy of the ModelParams.