Generate simulated data from the underlying model's generative process

Function that allows the user to generate hospital admissions and site-level wastewater data directly from the generative model, specifying the conditions and parameters to generate from.

Usage

generate_simulated_data(
  r_in_weeks = c(rep(1.1, 5), rep(0.9, 5), 1 + 0.007 * 1:16),
  n_sites = 4,
  ww_pop_sites = c(4e+05, 2e+05, 1e+05, 50000),
  pop_size = 3e+06,
  site = c(1, 1, 2, 3, 4),
  lab = c(1, 2, 3, 3, 3),
  ot = 90,
  nt = 9,
  forecast_horizon = 28,
  sim_start_date = lubridate::ymd("2023-09-01"),
  hosp_wday_effect = c(0.95, 1.01, 1.02, 1.02, 1.01, 1, 0.99)/7,
  i0_over_n = 5e-04,
  initial_growth = 1e-04,
  sd_in_lab_level_multiplier = 0.25,
  mean_obs_error_in_ww_lab_site = 0.2,
  mean_reporting_freq = 1/5,
  sd_reporting_freq = 1/20,
  mean_reporting_latency = 7,
  sd_reporting_latency = 3,
  mean_log_lod = 5,
  sd_log_lod = 0.2,
  global_rt_sd = 0.03,
  sigma_eps = 0.05,
  sd_i0_over_n = 0.5,
  if_feedback = FALSE,
  subpop_phi = c(25, 50, 70, 40, 100),
  input_params_path = fs::path_package("extdata", "example_params.toml", package =
    "wwinference")
)

Arguments

r_in_weeks: vector indcating the mean weekly R(t) that drives infection dynamics at the state-level. This gets jittered with random noise to add week-to-week variation.
n_sites: integer indicating the number of sites
ww_pop_sites: vector indicating the population size in the catchment area in each of those sites (order must match)
pop_size: integer indicating the population size in the hypothetical state, default is 3e6
site: vector of integers indicating which site (WWTP) each separate lab-site observation comes from
lab: vector of integers indicating which lab the lab-site observations come from
ot: integer indicating the observed time: length of hospital admissions calibration time in days
nt: integer indicating the nowcast time: length of time between last hospital admissions date and forecast date in days
forecast_horizon: integer indicating the duration of the forecast in days e.g. 28 days
sim_start_date: character string in ISO8601 format YYYY-MM-DD indicating the start date of the simulation, used to get a weekday vector
hosp_wday_effect: a vector that is a simplex of length 7 describing how the hospital admissions are spread out over a week, starting at Monday = 1
i0_over_n: float between 0 and 1 indicating the initial per capita infections in the state
initial_growth: float indicating the exponential growth rate in infections (daily) during the unobserved time
sd_in_lab_level_multiplier: float indicating the standard deviation in the log of the site-lab level multiplier determining how much variation there is systematically in site-labs from the state mean
mean_obs_error_in_ww_lab_site: float indicating the mean day-to-day variation in observed wastewater concentrations across all lab-sites
mean_reporting_freq: float indicating the mean frequency of wastewater measurements across sites in per day (e.g. 1/7 is once per week)
sd_reporting_freq: float indicating the standard deviation in the frequency of wastewater measurements across sites
mean_reporting_latency: float indicating the mean time from forecast date to last wastewater sample collection date, across sites
sd_reporting_latency: float indicating the standard deviation in the time from the forecast date to the last wastewater sample collection date, across sites
mean_log_lod: float indicating the mean log of the LOD in each lab-site
sd_log_lod: float indicating the standard deviation in the log of the LOD across sites
global_rt_sd: float indicating the ammount of standard deviation to add to the passed in weekly R(t) to add variability. Default is 0.03
sigma_eps: float indicating the standard deviation between the log of the state R(t) and the log of the subpopulation R(t) across time, in log scale. Default is 0.05
sd_i0_over_n: float indicating the standard deviation between log of initial infections per capita, default is 0.5
if_feedback: Boolean indicating whether or not to include infection feedback into the infection process, default is FALSE, which sets the strength of the infection feedback to 0. If TRUE, this will apply an infection feedback drawn from the prior.
subpop_phi: Vector of numeric values indicating the overdispersion parameter phi in the hospital admissions observation process in each subpopulation
input_params_path: path to the toml file with the parameters to use to generate the simulated data

Value

a list containing three dataframes. hosp_data is a dataframe containing the number of daily hospital admissions by day for a theoretical US state, for the duration of the specified calibration period. hosp_data_eval is a dataframe containing the number of daily hospital admissions by day for a theoretical US state, for the entire evaluation period. ww_data is a dataframe containing the measured wastewater concentrations in each site alongside other metadata necessary for modeling that data.

Examples

if (FALSE) { # \dontrun{
# Generate a simulated dataset from a hypothetical state with 6 sites and 2
# different labs
sim_data <- generate_simulated_data(
  n_sites = 6,
  site = c(1, 2, 3, 4, 5, 6, 6),
  lab = c(1, 1, 1, 1, 2, 2, 3),
  ww_pop_sites = c(1e5, 4e5, 2e5, 1.5e5, 5e4, 3e5),
  pop_size = 2e6
)
hosp_data <- sim_data$hosp_data
ww_data <- sim_data$ww_data
} # }