{
 "version":87,
 "stage":"replicate_weight_informed_precision_and_outcome_composition_membership_audit",
 "origin_panels":[2018,2020,2021,2022,2023],
 "primary_domain":"Exact v85 strict fixed-index household membership. Reconstruction must reproduce all cohort cell counts.",
 "sensitivity_domain":"Relax ONLY the outcome-year exact co-resident roster rule. Retain event-year exact roster, fixed baseline index, any-baseline-member event exposure, baseline and outcome all-12-month index residence at interview address, positive same endpoint FINYR3 and both-period v85 validated food coding. Outcome-year entrants/departures permitted; no claim to cover event-year splits or movers.",
 "food_use":"Only v85 required-path and recode-consistency validity indicators. No hardship scores, outcomes, means or contrasts by exposure/resources are emitted or fitted.",
 "primary_precision":"Within each domain/cohort, merge same index point and 240 longitudinal replicate weights; verify FINYR3 type, exact reference years and REPWGT0 agreement; derive the linear four-cell change contrast matrix. Working-model E[Fay variance] uses independent household changes of common variance; this is not an observed-outcome survey SE.",
 "fay_replicates":240,
 "fay_factor":0.5,
 "cohort_aggregation":"Equal five-cohort planning. Give independent-cohort sum(C)/25 and perfect-positive-covariance envelope (sum(sqrt(C))/5)^2. No pairing of unrelated replicate columns across endpoints.",
 "change_variance_scenarios":{"prevalence":[0.05,0.1,0.2,0.5],"prepost_correlation":[0,0.5,0.8]},
 "halfwidth_target_pp":5,
 "selection_outputs":"Sequential outcome-residence, outcome-roster and food-quality exclusions by cohort/exposure/deposits. Descriptive point-weight retention is index-person-domain weighted, not national household prevalence. No selection correction is fit.",
 "effect_estimation":false,
 "actual_outcome_standard_errors":false,
 "model_calibration":false,
 "known_before_plan":"v85/v86 aggregate cell and flow counts; prior person-level v82/v83 outcome results. No new household liquidity-outcome contrast has been inspected. Not external preregistration."
}