Parameters

class pypomp.core.parameters.PompParameters(theta: Mapping[str, int | float | number | Array] | Sequence[Mapping[str, int | float | number | Array]] | PompParameters | DataArray | None, logLik: ndarray | None = None, estimation_scale: bool = False)

Bases: ParameterSet[DataArray]

Manages parameters for a standard Pomp model.

Internal storage is a 3D xarray.DataArray with dimensions ("theta_idx", "unit", "parameter"), where "unit" is always "shared".

Parameters:

• theta (Mapping[str, Numeric] | Sequence[Mapping[str, Numeric]] | PompParameters | xr.DataArray | None) –

  Parameters for the model. Accepts:

  • A single dictionary: dict[str, Numeric]

  • A list of dictionaries: list[dict[str, Numeric]]

  • An existing PompParameters object

  • An xarray.DataArray with dimensions ("theta_idx", "unit", "parameter")

• logLik (np.ndarray, optional) – A numpy array of log-likelihoods associated with each parameter set.

• estimation_scale (bool, optional) – Whether the parameters are on the estimation scale. Defaults to False.

Methods

logLik	Get or set the log-likelihoods for each parameter set (theta_idx).
to_jax_array([param_names])	Convert to a JAX array matching the order of param_names.
num_replicates()	Returns the number of parameter sets/replicates.
num_params()	Return the number of canonical parameters.
subset(indices)	Return a new PompParameters object with the specified parameter set (theta_idx) indices.
get_param_names()	Return the list of parameter names contained in this set.
params(-> list[dict[str, ...)	Get the parameters in this set.
set_params(value)	Set or overwrite the parameter values.
transform(par_trans[, direction])	Transform the parameters to or from the estimation parameter space.
prune([n, refill])	Replace internal parameter sets with the top n based on stored log-likelihoods.
merge(*param_objs)	Merge replications from an arbitrary number of PompParameters objects.

class pypomp.core.parameters.PanelParameters(theta: Mapping[str, DataFrame | None] | Sequence[Mapping[str, DataFrame | None]] | PanelParameters | Dataset | None, logLik_unit: ndarray | None = None, estimation_scale: bool = False)

Bases: ParameterSet[Dataset]

Manages parameters for PanelPomp models.

Internal storage is a 3D xarray.DataArray with dimensions ("theta_idx", "unit", "parameter").

Parameters:

• theta (PanelParameters | dict | list | xr.DataArray, optional) –

  Parameters for the panel model. Accepts:

  • A single dictionary with "shared" and "unit_specific" keys (each containing a DataFrame).

  • A list of such dictionaries.

  • An existing PanelParameters object.

  • An existing xarray.DataArray with dimensions ("theta_idx", "unit", "parameter").

• logLik_unit (np.ndarray, optional) – A numpy array of unit-specific log-likelihoods of shape (num_theta_idx, n_units).

• estimation_scale (bool, optional) – Whether the parameters are on the estimation scale. Defaults to False.

Methods

logLik	Get the overall log-likelihood for each parameter set (theta_idx).
logLik_unit	Get or set the unit-specific log-likelihoods for each parameter set (theta_idx).
params(-> list[dict[str, ...)	Get the parameters in this set.
set_params(value)	Set or overwrite the parameter values.
num_replicates()	Returns the number of parameter sets/replicates.
get_param_names()	Return the list of parameter names contained in this set.
get_shared_param_names()	Return the list of shared parameter names.
get_unit_param_names()	Return the list of unit-specific parameter names.
get_unit_names()	Return the list of unit names.
subset(indices)	Return a new PanelParameters object with the specified parameter set (theta_idx) indices.
to_jax_array([param_names, unit_names])	Convert to a JAX array matching the order of param_names and unit_names.
transform(par_trans[, direction])	Transform the parameters to or from the estimation parameter space.
prune([n, refill])	Replace internal parameter sets with the top n based on stored log-likelihoods.
mix_and_match()	Mixes unit-specific and shared parameters independently by sorting each unit's unit-specific parameters and the shared parameters in descending order of their respective log-likelihood contribution.