Skip to contents

Default Simulation Configuration

Source: R/config_default.R
config_default.Rd

The canonical simulation parameter object shipped with MOSAIC. It holds default values for every model parameter, the initial state vectors, the daily input matrices and the observed surveillance series for the 40 modelled locations, and is the template that get_location_config() subsets and sample_parameters() draws into.

Usage

config_default

Format

A named list built by the data-raw/make_config_default.R script and persisted as data/config_default.rda (its version is in metadata$version). Element groups:

  • Metadata – metadata$version, metadata$date and metadata$description (the full change log) for provenance tracking;

  • Simulation window and locations – seed, date_start, date_stop, location_name, longitude, latitude;

  • Initial conditions – counts (N_j_initial, S_j_initial, E_j_initial, I_j_initial, R_j_initial, V1_j_initial, V2_j_initial) and the matching proportions (prop_*_initial);

  • Global scalars – disease and immunity rates (iota, gamma_1, gamma_2, epsilon, phi_1, phi_2, omega_1, omega_2), the observation process (rho, rho_deaths, sigma, chi_endemic, chi_epidemic, delta_reporting_cases), mobility (mobility_omega, mobility_gamma), mixing (alpha_2), shedding and dose (zeta_1, zeta_2, zeta_ratio, kappa), environmental decay (decay_days_short, decay_days_spread, decay_days_long, decay_shape_1, decay_shape_2) and the seasonal period p;

  • Per-location vectors – transmission (beta_j0_tot, p_beta and the derived beta_j0_hum, beta_j0_env), alpha_1, tau_i, theta_j, seasonality (a_1_j, a_2_j, b_1_j, b_2_j), epidemic_threshold and the suitability calibration psi_star_a, psi_star_b, psi_star_z, psi_star_k;

  • Location x day matrices – birth and death rates (b_jt, d_jt), first- and second-dose OCV doses per day (nu_1_jt, nu_2_jt, with the compartments eligible for first doses in nu_jt_sources; see Details), environmental suitability (psi_jt, uncalibrated: sample_parameters() applies psi_star_* to it), and the reported case fatality ratio mu_jt, the deaths input of the mortality model since MOSAIC v0.96.0;

  • Observed data – reported_cases and reported_deaths; their per-observation confidence weights reported_cases_weight and reported_deaths_weight; reported_tier (since MOSAIC v0.101.0), the surveillance trust tier of each observed day's week (1 observed, 2 reconstructed: a WHO multi-week report spread over its weeks, 3 imputed; NA where neither cases nor deaths are observed), which the dispersion estimates read and the engine ignores; and the epidemic_peaks data frame used by the peak shape terms of the likelihood. The cases dispersion is fitted on tier-1 weeks only; the deaths dispersion is fitted on tier-1 weeks where they are enough and falls back to every scored week, tiers 2 and 3 included, where they are too few (for the integrated deaths likelihood on v7.0 at burn_in_days = 45: BEN, BFA, CIV, LBR, MLI, NAM and ZAF).

Details

The object is self-contained: it references no external files and runs directly, e.g. run_simulation(config_default). The engine does not read psi_star_*, so a direct run uses the raw psi_jt and ignores the shipped psi_star_b = 1; to simulate with the calibrated suitability, pass the config through sample_parameters() first. For tutorials or fast tests the smaller toy configs config_simulation_epidemic and config_simulation_endemic are also available.

Vaccination doses: nu_1_jt + nu_2_jt is the OCV doses shipped to each location, administered at up to 20,000 doses a day by est_vaccination_rate() from the GTFCC/WHO campaign file, each GTFCC delivery from its own delivery date. Since config_default v7.0 (MOSAIC v0.103.0) each request's doses are split by its GTFCC rounds: shipped doses are divided across the rounds in proportion to the doses administered in each, the first round goes to nu_1_jt and the second to nu_2_jt, in that order, and doses with no round information (a request without GTFCC Round events, or a WHO-only shipment) are first doses. The engine moves phi_1 of each day's first doses from the nu_jt_sources compartments to V1, and phi_2 of its second doses from V1 to V2, capped at the V1 stock, so a two-dose campaign immunises its first-round recipients once. Second rounds are pre-2023 campaigns (the ICG suspended two-dose outbreak response in October 2022), so over a 2023+ window nu_2_jt is zero. Up to v6.2 every dose was a first dose, nu_2_jt was zero everywhere, and every delivery of a request started at its first delivery date (this moved later deliveries of 22 pre-2023 requests up to 2.6 years (961 days) early and did not change the 2023+ series).

Note on Initial Condition Formats: This configuration includes both count (*_j_initial) and proportion (prop_*_initial) representations of initial conditions. The count fields are required by the transmission model for simulation, while the proportion fields are optional and provided for statistical analysis convenience. Both formats are automatically maintained in sync during parameter sampling operations.

See also