Quick Start

This page shows the public CCP wrapper. The snippet checks that the estimator runs and returns the structural objects; support and first-stage CCP quality are checked on the pre-estimation page.

The public API follows the same sklearn convention as NFXP. Create an estimator, call fit, and read fitted attributes.

from econirl.datasets import load_rust_bus
from econirl import CCP

df = load_rust_bus()

model = CCP(
    n_states=90,
    discount=0.9999,
    utility="linear_cost",
    num_policy_iterations=10,
)
model.fit(df, state="mileage_bin", action="replaced", id="bus_id")

print(model.params_)
print(model.se_)
print(model.policy_.shape)

Output

{'theta_c': 0.0010028827236965299, 'RC': 3.072209358215332}
{'theta_c': 0.00038928037896014613, 'RC': 0.07406123281287358}
(90, 2)

The fitted estimator exposes structural parameters, standard errors, a policy, a value function, and a likelihood.

Attribute	Meaning
params_	Estimated structural reward parameters.
se_	Standard errors for the structural parameters.
policy_	Estimated action probabilities by state.
value_	Estimated policy value function by state.
log_likelihood_	Maximized CCP pseudo log likelihood.

one_step = CCP(n_states=90, discount=0.9999, num_policy_iterations=1)
npl = CCP(n_states=90, discount=0.9999, num_policy_iterations=10)

The one-step model uses the empirical CCPs once. The NPL model updates the policy after each pseudo-likelihood fit.

Counterfactual Example

cf = model.counterfactual(RC=4.0)

print(cf.params)
print(cf.policy[50, 1])

Output

{'theta_c': 0.0010028827236965299, 'RC': 4.0}
0.055194772664901845

This solves the fitted model again with a higher replacement cost and returns the new value function and policy.

Full Estimator API

Use econirl.estimation.ccp.CCPEstimator when you need direct control over panel objects, utility objects, transition tensors, CCP smoothing, NPL stopping rules, or diagnostic metadata.