ShawnChamberlain's picture
Add files using upload-large-folder tool
2ed92a1 verified
Raw
History Blame Contribute Delete
2.02 kB
{
"assignment_design": "balanced geographic cluster saturation arms followed by independent within-zone-period opportunity assignment",
"config": {
"cluster_noise_sd": 0.04,
"confidence_level": 0.95,
"history_effect": 0.7,
"history_lags": 1,
"individuals_per_cell": 20,
"minimum_inference_clusters": 8,
"n_clusters": 32,
"n_periods": 32,
"n_zones": 32,
"neighbor_effect": 1.5,
"outcome_noise_sd": 0.08,
"own_effect": 2.0,
"replications": 24,
"saturation_levels": [
0.0,
0.3333333333333333,
0.6666666666666666,
1.0
],
"seed": 880321
},
"estimand_definitions": {
"controlled_history_exposure_response": "Per-unit response to the exact-lag own-treatment history holding current own and mapped-neighbor exposure fixed.",
"controlled_zone_direct_effect": "Per-unit own-treatment response holding mapped neighbor and lagged exposure fixed.",
"market_total_effect": "Full-horizon all-zone treatment versus all-zero under the benchmark's additive exposure mapping and startup history convention.",
"spillover_effect": "Per-unit response to the predeclared mapped-neighbor exposure holding own and lagged exposure fixed."
},
"evidence_type": "semi_synthetic_exposure_mapped_known_truth_benchmark",
"exposure_mapping": "two equal-weight ring neighbors declared before outcomes",
"history_mapping": "mean of 1 exact own-treatment lag(s)",
"known_estimands": {
"controlled_history_exposure_response": 0.7,
"controlled_zone_direct_effect": 2.0,
"full_horizon_persistent_effect": 0.678125,
"market_total_effect": 4.178125,
"spillover_effect": 1.5
},
"market_total_bridge": "Known only because this benchmark declares an additive exposure-response DGP; controlled coefficients are not individually relabeled as market_total_effect.",
"naive_assignment_status": "target mismatch; coefficient uncertainty is cluster-aware, but market-total bias, RMSE, coverage, and power are withheld"
}