robynstuart/hivsim_zim

HIVsim Zimbabwe model # hivsim_zim HIVsim Zimbabwe: a calibrated HIV model on the Starsim / STIsim stack, used to reproduce published Zimbabwe results from two multi-model comparison studies and to draft a validation preprint. See `ANALYSIS_PLAN.md` for the sprint brief. ## What's in the repo - `model.py` — HIV-only Zimbabwe sim (HIV disease + structured sexual network + Zimbabwe demographics), 1985–2040. Slim wrapper around stisim's `sti.HIV` and `sti.StructuredSexual`. - `hiv_model.py` — HIV disease module + interventions (testing programs, ART, VMMC, PrEP) with time-varying coverage from `data/n_art.csv`, `data/n_vmmc.csv`. - `priors.py` — HIV-relevant calibration priors (`hiv.beta_m2f`, `hiv.rel_init_prev`, network shape). - `data/` — Zimbabwe HIV surveillance (`zimbabwe_hiv_calib.csv`), initial prevalence, ART / VMMC coverage, demographics (age structure, ASFR, deaths, migration, condom use). - `calibration/artifacts/` — 500-draw LHS × K=5 sim-averaging calibration outputs from the `sti_notification` project (experiment 06, 2026-06-24). Includes the top-10 draws by GoF (`draws_top10.csv`), the full 500-draw prior sample (`priors_500.csv`), per-draw calibration metrics (`per_draw_means_wholepop.csv`), and the calibration write-up (`CALIBRATION_SUMMARY.md`). - `reference/` — digitised / appendix data from the target papers (populated during the sprint). - `outputs/` — simulated indicator time series (gitignored). - `figures/` — comparison figures (gitignored). ## Provenance The calibrated model is lifted from sti_notification (private, IDM), where HIV was calibrated jointly with syphilis + NG/CT/TV/BV against Zimbabwe MoH and ZIMPHIA data. The joint HIV–syph coupling in that fit is `hiv → syph` (HIV+ agents are more susceptible to syph), not the reverse, so dropping the STIs here does not materially perturb the HIV trajectory. The calibration itself ran on stisim `fix/ng-tx@731bc1d`. ## Quick start ```bash pip install -r requirements.txt python model.py # smoke t …