Skip to content

Reporting

These parameters govern how the model's internal transitions (new symptomatic infections and disease deaths) are turned into observable time series, and how those observations are aligned against external data. The first four — rho, rho_deaths, delta_reporting_cases, delta_reporting_deaths — are scalar knobs that thin and lag the simulated streams, producing the reported_cases / reported_deaths outputs the analyzer scores. The last two — reported_cases and reported_deaths — are the observed-data arrays themselves: per-patch, per-time-step targets that the likelihood scorer compares the simulation against.

Together they form the bridge between the underlying SEIR dynamics in infectious.py and the likelihood machinery in analyzer.py. With all four scalar knobs at their off-values (rho = 1, rho_deaths = 1, delta_reporting_cases = 0, delta_reporting_deaths = 0) the reported series become same-tick, fully-observed copies of the internal counts (subject only to the regime-switching scaling described in Regime switching).

Quick reference

Parameter Shape Dtype Range Off-value
rho scalar float32 [0, 1] 1.0
rho_deaths scalar float32 [0, 1] 1.0
delta_reporting_cases scalar int32 >= 0 0
delta_reporting_deaths scalar int32 >= 0 0
reported_cases (npatches, n_obs_timesteps) float32 shape-only check n/a (not a feature toggle)
reported_deaths (npatches, n_obs_timesteps) float32 shape-only check n/a (not a feature toggle)

rho

  • What it controls: Per-case reporting probability for symptomatic cases — fraction of new symptomatic infections that show up in the reported_cases series.
  • Shape: scalar
  • Dtype: float32
  • Range: [0, 1]
  • Off-value: 1.0 — every new symptomatic case is counted (the Binomial(n, 1) == n identity), leaving the regime-switching scaling as the only remaining transform between internal new_symptomatic and the reported series.
  • Consumer code: src/laser/cholera/metapop/infectious.py
  • Related parameters: rho_deaths, delta_reporting_cases, chi_endemic, chi_epidemic, epidemic_threshold, reported_cases
  • Notes: infectious.py:267 draws Binomial(new_symptomatic, rho) then divides by chi_eff; with rho = 1 every symptomatic case is counted, so the only remaining transform is the chi_endemic / chi_epidemic scaling. To get strict "observed == estimated symptomatic" also requires chi_endemic == chi_epidemic == 1.

rho_deaths

  • What it controls: Per-death reporting probability — fraction of disease deaths that show up in the reported_deaths series.
  • Shape: scalar
  • Dtype: float32
  • Range: [0, 1]
  • Off-value: 1.0 — every disease death is counted, so the reported-deaths series equals the simulated disease-death count (up to the delta_reporting_deaths lag).
  • Consumer code: src/laser/cholera/metapop/infectious.py
  • Related parameters: rho, delta_reporting_deaths, reported_deaths

delta_reporting_cases

  • What it controls: Integer tick lag between when a new symptomatic case occurs and when it is added to reported_cases.
  • Shape: scalar
  • Dtype: int32
  • Range: >= 0
  • Off-value: 0 — cases are written into the reported series on the same tick they occur (no lag).
  • Consumer code: src/laser/cholera/metapop/infectious.py
  • Related parameters: rho, delta_reporting_deaths, epidemic_threshold
  • Notes: Also used at infectious.py:196 to compute the epidemic_flag from a lagged Isym snapshot, so this parameter also shifts the regime-switching probe — not just the reporting series.

delta_reporting_deaths

  • What it controls: Integer tick lag between when a disease death occurs and when it is added to reported_deaths.
  • Shape: scalar
  • Dtype: int32
  • Range: >= 0
  • Off-value: 0 — deaths are written into the reported series on the same tick they occur (no lag).
  • Consumer code: src/laser/cholera/metapop/infectious.py
  • Related parameters: rho_deaths, delta_reporting_cases

reported_cases

  • What it controls: Observed per-patch-per-time-step case counts used as the target series when computing the model likelihood.
  • Shape: (npatches, n_obs_timesteps)
  • Dtype: float32
  • Range: shape-only check (type assert: must be list or ndarray; NaN allowed for missing weeks).
  • Off-value: n/a (not a feature toggle) — reported_cases is observed data, not a switch. To run the model without scoring against case data, set calc_likelihood to False instead; the analyzer only consumes reported_cases when calc_likelihood is on.
  • Consumer code: src/laser/cholera/metapop/analyzer.py
  • Related parameters: reported_deaths, calc_likelihood, rho, delta_reporting_cases, weight_cases
  • Notes: Ingested via handle_nan, which converts non-numeric cells to np.nan (hence float32, not int). The only invariant enforced is type (list-of-lists or ndarray) in dict_to_propertysetex — no shape or range check in validate_parameters. The analyzer slices by reported_cases.shape[1] columns, confirming the (npatches, n_obs_timesteps) layout.

reported_deaths

  • What it controls: Observed per-patch-per-time-step death counts used as the target series when computing the model likelihood.
  • Shape: (npatches, n_obs_timesteps)
  • Dtype: float32
  • Range: shape-only check (type assert: must be list or ndarray; NaN allowed for missing weeks).
  • Off-value: n/a (not a feature toggle) — reported_deaths is observed data, not a switch. To run the model without scoring against death data, set calc_likelihood to False instead; the analyzer only consumes reported_deaths when calc_likelihood is on.
  • Consumer code: src/laser/cholera/metapop/analyzer.py
  • Related parameters: reported_cases, calc_likelihood, rho_deaths, delta_reporting_deaths, weight_deaths
  • Notes: Same ingestion path as reported_cases (handle_nan -> float32 to accommodate NaN missing cells). No shape or range invariant in validate_parameters; only the type assert in dict_to_propertysetex. Consumed only by the analyzer when calc_likelihood is True.