{ "version": "partition-bootstrap-t-posthoc-1.0", "status": "post_hoc_conditional_calibration_diagnostic_no_promotion", "protocol": { "version": "partition-bootstrap-t-posthoc-1.0", "status": "post_hoc_conditional_calibration_diagnostic_no_promotion", "registration": "Fixed after observing scale-refit partition interval undercoverage, before running this nested bootstrap experiment. Baseline and temporal-nugget grids and outputs remain unchanged; no setting is selected for significance or favorable coverage.", "index": "ndre", "scenarios": [ { "tau_days": 30, "range_m": 150, "common_fraction": 0.25, "temporal_nugget": 0 }, { "tau_days": 30, "range_m": 150, "common_fraction": 0.25, "temporal_nugget": 0.25 } ], "outer_replicates": 200, "inner_replicates": 399, "observed_bootstrap_replicates": 399, "seed": 2026100481, "partition_seed": 20261004, "confidence_level": 0.95, "estimation_target": "Shared temporal slope under the common-slope working model, estimated by averaging four within-partition GLS coefficients", "notes": [ "This target differs from the main finite-map arithmetic mean when cell slopes are heterogeneous. The arithmetic mean results remain the baseline comparator and are not replaced.", "Each outer Gaussian dataset is independently generated with the observed-data residual scales and fixed covariance parameters. Its temporal coefficients and residual scales are refitted, then its partition weights and cross-partition covariance are recomputed.", "Every outer dataset runs its own independent inner parametric bootstrap using that outer dataset's estimated scales. Inner temporal coefficients, residual scales and partition weights are all refitted. No outer outcome selects or tunes a critical value.", "Inner data use the common-slope null zero; location equivariance makes this equivalent to generating at the fitted common slope and centering studentized errors. Intercepts and July terms lie in the fitted mean-model span and can be set to zero for this calculation.", "The symmetric bootstrap-t interval uses q from absolute studentized errors |theta_star|/SE_star; one-based quantile rank is ceil((B+1)*0.95). This rank convention is not a conformal or finite-sample coverage guarantee.", "Binomial order-statistic bounds describe Monte Carlo uncertainty in the fitted bootstrap distribution's 95th percentile. Wilson bounds describe finite-outer-simulation coverage uncertainty; neither assesses real forest accuracy.", "The observed-data bootstrap interval is a retrospective common-slope working-model diagnostic. Covariance misspecification, heterogeneous trends, support selection and ecological uncertainty remain unvalidated even if null coverage improves.", "Normal and bootstrap coverage use the same outer datasets for a paired comparison. Outer/observed/inner streams use separate deterministic seed vectors.", "No production confidence layer, p-value, forest-damage probability, detector retuning or original-data modification is authorized by this diagnostic." ] }, "frozen_file_sha256": { "C:\\badgr\\coordinate_forest_ml_webapp\\experiments\\parts_inference\\bootstrap_t\\protocol.json": "906c50a34c1474bf0569b65ceaec8732551a9fb690136b6483ea81d690ae0564", "C:\\badgr\\coordinate_forest_ml_webapp\\experiments\\parts_inference\\bootstrap_t\\bootstrap_t.py": "6b50a8ec2aac6c452446755885ea7ae2a7e9d8b5858b2048a64971ac0a18ccd5", "C:\\badgr\\coordinate_forest_ml_webapp\\experiments\\parts_inference\\measurement_noise\\baseline_snapshot\\protocol.json": "5debfd356920933316cc2ba0219fb28eab7d45b023b265aeb2bb4f3c43665be6", "C:\\badgr\\coordinate_forest_ml_webapp\\experiments\\parts_inference\\measurement_noise\\measurement_noise.py": "99cff60551c4b9365add53ec696bd4ef6be9c117439dda7ba302d25795b315d2", "C:\\badgr\\coordinate_forest_ml_webapp\\experiments\\parts_inference\\measurement_noise\\results\\summary.json": "8b269d5680424f163ab8687b4738bd161b8a2c60b6a6f3b8fad521b0bb55b328" }, "inputs": { "observations": { "path": "C:\\badgr\\coordinate_forest_ml_webapp\\runtime\\observation_audit_v1\\cell_date_observations.csv", "sha256": "1a509ecd85a1d8810d5ef35d55e1760cc0b63a627ecefea1ed843b8ee55958de" }, "cells": { "path": "C:\\badgr\\coordinate_forest_ml_webapp\\runtime\\pilot_v4\\tables\\cell_scores.csv", "sha256": "2fb18f884be7bbe80f1ad90ee1abe24c81ef2523061dd0723ea6ed7ddb8aa65e" } }, "source_and_baseline_unchanged": true, "observed_intervals": [ { "index": "ndre", "case": 0, "tau_days": 30, "range_m": 150, "common_fraction": 0.25, "temporal_nugget": 0, "estimate": 0.0053051307202723426, "plugin_standard_error": 0.022526506585838633, "normal_interval_lower": -0.038846010885475714, "normal_interval_upper": 0.049456272326020394, "bootstrap_interval_lower": -0.06477744147974267, "bootstrap_interval_upper": 0.07538770292028736, "bootstrap_to_normal_width_ratio": 1.5873331844015317, "critical_value": 3.1111158728922783, "critical_rank": 380, "critical_mc_lower": 2.8183466886413684, "critical_mc_upper": 3.669722591434702, "critical_mc_lower_rank": 370, "critical_mc_upper_rank": 388 }, { "index": "ndre", "case": 1, "tau_days": 30, "range_m": 150, "common_fraction": 0.25, "temporal_nugget": 0.25, "estimate": 0.00865105398952936, "plugin_standard_error": 0.007994101450119697, "normal_interval_lower": -0.007017096941464664, "normal_interval_upper": 0.024319204920523384, "bootstrap_interval_lower": -0.016145796510183715, "bootstrap_interval_upper": 0.03344790448924244, "bootstrap_to_normal_width_ratio": 1.582627752880595, "critical_value": 3.101893396579523, "critical_rank": 380, "critical_mc_lower": 2.777391688735127, "critical_mc_upper": 3.52022365971631, "critical_mc_lower_rank": 370, "critical_mc_upper_rank": 388 } ], "outer_validation": [ { "index": "ndre", "case": 0, "tau_days": 30, "range_m": 150, "common_fraction": 0.25, "temporal_nugget": 0, "outer_replicates": 200, "inner_replicates_each": 399, "normal": { "coverage": 0.715, "mc_standard_error": 0.03191982142807193, "mc_wilson_lower": 0.6488465397039482, "mc_wilson_upper": 0.7730499699538271 }, "bootstrap_t": { "coverage": 0.95, "mc_standard_error": 0.015411035007422448, "mc_wilson_lower": 0.9104218518612239, "mc_wilson_upper": 0.972617354399236 }, "paired_coverage_gain": 0.235, "paired_gain_monte_carlo_se": 0.03005647949775547, "normal_miss_bootstrap_cover": 47, "normal_cover_bootstrap_miss": 0, "mean_width_ratio": 1.59337686018539, "median_critical_value": 3.114345116671153 }, { "index": "ndre", "case": 1, "tau_days": 30, "range_m": 150, "common_fraction": 0.25, "temporal_nugget": 0.25, "outer_replicates": 200, "inner_replicates_each": 399, "normal": { "coverage": 0.785, "mc_standard_error": 0.02904952667428507, "mc_wilson_lower": 0.7229769265100339, "mc_wilson_upper": 0.8362812374549242 }, "bootstrap_t": { "coverage": 0.92, "mc_standard_error": 0.019183326093250873, "mc_wilson_lower": 0.874010511732856, "mc_wilson_upper": 0.9501594141102401 }, "paired_coverage_gain": 0.135, "paired_gain_monte_carlo_se": 0.024224141785131024, "normal_miss_bootstrap_cover": 27, "normal_cover_bootstrap_miss": 0, "mean_width_ratio": 1.5939827218513967, "median_critical_value": 3.1205141152856095 } ], "no_production_promotion": true, "output_sha256": { "observed_intervals.csv": "008966119e52b0d1ae4ef11fedea082dc4a0e0e7ebdc42c11599fed90c62db8e", "outer_validation.csv": "2042cb40477b2e43a88be93611d034307f41a162d68bee8ea4d01a51b1dd9c2d", "observed_bootstrap_draws.csv": "77602505c02801a9e4234ea5fa297301d9ebc49b3f522d6374a12fd83d012698" }, "interpretation": "Paired nested null validation measures conditional calibration under a fixed common-slope Gaussian covariance model. It does not validate the covariance model or heterogeneous finite-map, ecological or prospective confidence. Observed intervals are retrospective working-model diagnostics only." }