PanelPomp Class

class pypomp.PanelPomp(Pomp_dict: dict[str, Pomp], theta: PanelParameters)

Bases: PanelValidationMixin, PanelEstimationMixin, PanelAnalysisMixin

Panel of partially observed Markov process models.

Extends the single-unit POMP framework to handle multiple units that share structural characteristics but may have distinct parameter values and observations.

Parameters:

• Pomp_dict (dict of str to Pomp) – Mapping from unit names to Pomp objects.

• theta (PanelParameters) – A PanelParameters object containing the model parameters.

Attributes

PanelPomp.ys

Tidy DataFrame containing observations from all units.

Returns:

Long-format DataFrame of observations.

Return type:

pd.DataFrame

PanelPomp.theta

The parameter object for the panel model.

PanelPomp.unit_objects: dict[str, Pomp]

A dictionary mapping unit names to their corresponding Pomp objects.

PanelPomp.canonical_param_names: list[str]

All unique parameter names present in either the shared or unit-specific parameters.

PanelPomp.canonical_shared_param_names: list[str]

Parameter names of parameters with values shared across all units in the panel.

PanelPomp.canonical_unit_param_names: list[str]

Parameter names of parameters with values specific to individual units in the panel.

PanelPomp.results_history: ResultsHistory

A ResultsHistory object storing the history of results from method calls.

PanelPomp.fresh_key: Array | None

Running a method that accepts a JAX PRNG key will store a fresh, unused key here.

PanelPomp.metadata: ModelMetadata

Environment and version metadata initialized when this instance was built.

Core Algorithmic Methods

simulate(key[, theta, times, nsim, as_pomp])	Simulate latent states and observations from the panel model.
pfilter(J[, key, theta, thresh, reps, ...])	Run the bootstrap particle filter (SMC) algorithm.
mif(J, M, rw_sd[, key, theta, thresh, ...])	Estimate parameters using the (Marginalized) Panel Iterated Filtering (MPIF/PIF) algorithm.
train(J, M, eta[, chunk_size, optimizer, ...])	Estimate parameters using MOP-based gradient-descent optimization.
dpop_train(J, M, eta[, chunk_size, ...])	Estimate parameters using DPOP-based gradient-descent optimization.

Results

results([index, ignore_nan])	Get the results DataFrame for the specified history index.
CLL([index, average])	Get conditional log-likelihoods for the specified history index.
ESS([index, average])	Get Effective Sample Size for the specified history index.
traces()	Get a tidy DataFrame with the full trace history of replicates.

Other Supporting Methods

get_unit_names()	Returns a list of the names of the units in the panel.
get_unit_parameters(unit[, theta])	Get parameter values for a specific unit across all replicates.
sample_params(param_bounds, units, n, key[, ...])	Sample parameters uniformly within bounds for a panel model.
to_struct()	Export static data and compiled simulators into a JAX PyTree.
prune([n, refill])	Prune replicates down to the top n by log-likelihood.
mix_and_match()	Sort parameters independently and cross-combine them.
time()	Get a DataFrame summarizing execution times of run methods.
probe(probes, key[, nsim, theta])	Evaluate probe statistics on real and simulated data for each unit.
arma([order, log_ys, suppress_warnings])	Fit an independent ARIMA model to the observation data of each unit.
negbin([autoregressive, suppress_warnings])	Fit a Negative Binomial model to the observation data of each unit.
print_metadata()	Prints the creation and runtime environment metadata for this instance.
print_summary([n])	Print a summary of the PanelPomp object.
merge(*panel_pomp_objs)	Merge replications from multiple PanelPomp objects into a single object.

Visualization

plot_traces([which, show])	Plot parameter and log-likelihood traces from the result history.
plot_simulations(key[, nsim, mode, theta, show])	Simulate and plot observations against real data.