inst/examples/germany_missing_report_dates.R

In epiforecasts/epinowcast: Flexible Hierarchical Nowcasting

# Load packages
library(epinowcast)
library(data.table)

# Set cmdstan path
cmdstanr::set_cmdstan_path()

# Use 2 cores
options(mc.cores = 2)

# Load and filter germany hospitalisations
nat_germany_hosp <- germany_covid19_hosp[location == "DE"][age_group == "00+"]
nat_germany_hosp <- enw_filter_report_dates(
  nat_germany_hosp,
  latest_date = "2021-10-01"
)

# Make sure observations are complete
nat_germany_hosp <- enw_complete_dates(
  nat_germany_hosp,
  by = c("location", "age_group")
)

# Make a retrospective dataset
retro_nat_germany <- enw_filter_report_dates(
  nat_germany_hosp,
  remove_days = 40
)
retro_nat_germany <- enw_filter_reference_dates(
  retro_nat_germany,
  include_days = 40
)

# Simulate missing data for a single reference date 
# We can't simulate missing data across reports because this would
# also require updating the known reporting framework
retro_nat_germany[reference_date == as.Date("2021-08-20"), confirm := NA]

# Add a flag for data not observed and impute for downstream preprocessing
retro_nat_germany <- retro_nat_germany |>
  enw_flag_observed_observations() |>
  enw_impute_na_observations()

# Get latest observations for the same time period
latest_obs <- enw_latest_data(nat_germany_hosp)
latest_obs <- enw_filter_reference_dates(
  latest_obs,
  remove_days = 40, include_days = 20
)

# Preprocess observations (note this maximum delay is likely too short)
pobs <- enw_preprocess_data(retro_nat_germany, max_delay = 20)

# Fit a simple nowcasting model with fixed growth rate and a
# log-normal reporting distribution.
nowcast <- epinowcast(pobs,
  expectation = enw_expectation(~1, data = pobs),
  fit = enw_fit_opts(
    save_warmup = FALSE, pp = TRUE,
    chains = 2, iter_warmup = 500, iter_sampling = 500,
  ),
  obs = enw_obs(
    family = "poisson", data = pobs, observation_indicator = ".observed"
  ),
)