Nanagriff/amr-lab-accreditation-misclassification

Monte Carlo simulation code and results for the study "Garbage in, policy out: laboratory accreditation status as a structured measurement-error mechanism in antimicrobial resistance estimates in sub-Saharan Africa." # Laboratory accreditation as a structured measurement-error mechanism in AMR estimates Monte Carlo simulation code and results for the study *"Garbage in, policy out: laboratory accreditation status as a structured measurement-error mechanism in antimicrobial resistance estimates in sub-Saharan Africa."* The study models laboratory accreditation status as a statistical parameter and quantifies how antimicrobial-susceptibility-testing (AST) misclassification propagates into estimated resistance prevalence, into the measured effect of an intervention, and into the smallest true change a surveillance network can reliably detect. A naive estimator (pooled proportion + Wilson interval) is compared with a misclassification-corrected Rogan–Gladen estimator across four accreditation profiles, three true prevalence levels, and two intervention scenarios. ## Contents | Path | Description | |---|---| | `amr_misclassification_sim.py` | Simulation engine (data-generating mechanism, both estimators, power/intervention scenarios). Fixed, recorded seeds. | | `results/simulation_results_summary.md` | Narrative summary of all results with the validation table. | | `results/results_coverage_5000rep.csv` | Coverage, bias, width, RMSE (+ Monte Carlo standard errors) — 5,000 reps, 12 profile×prevalence cells. | | `results/results_factorial.csv` | Full factorial: prevalence × profile × isolates (30/100/300) × labs (5/20/50). | | `results/results_bias_curve.csv` | Analytic naive bias vs prevalence, per profile. | | `results/results_crossovers.csv` | Weighted Se/Sp and closed-form bias-crossover prevalence, per profile. | | `results/results_power_frontier.csv` | Power to detect a reduction vs true effect size, per profile. | | `results/results_sensitivity_tiermapping.csv` | Tier-to-error mapping sensitivity analysis. | | `results/results_interventions.csv` | Non-differential and differential intervention scenarios. | | `figures/fig1_coverage.png` | 95% interval coverage by profile at t …