Wrapper function to take a location's training data and fit to the site-level model
Usage
fit_site_level_model(
train_data,
params,
model_file,
forecast_date,
forecast_time,
generation_interval,
inf_to_hosp,
infection_feedback_pmf,
model_type,
output_dir,
adapt_delta,
max_treedepth,
seed,
include_hosp = 1,
compute_likelihood = 1,
n_draws = 100,
n_chains = 4,
iter_sampling = 500,
iter_warmup = 250,
n_parallel_chains = 4,
write_files = TRUE,
output_full_df = FALSE,
...
)Arguments
- train_data
dataframe containing the hospital admissions on each day, the wastewater concentrations in each site and lab, and other metadata needed to pass to stan to run the model
- params
disease-specific parameters, including hyperparameters for priors.
- model_file
the compiled stan model
- forecast_date
date of the forecast
- forecast_time
duration of the forecast period
- generation_interval
a discretized (in days) normalized pmf indexed starting at 1 of the probability of each time from initial infection to secondary transmission
- inf_to_hosp
a discretized (in days) normalized pmf indexed starting at 0 of the probability of time from initial infection to hospital admissions
- infection_feedback_pmf
a discretized (in days) normalized pmf that dictates the distribution of delays from incident infection to incidence feedback
- model_type
name of the model being fit. Options are
"site-level infection dynamics","site-level observation error","site-level time-varying concentration"- output_dir
location to save the model outputs
- adapt_delta
dapt_delta target proposal acceptance probability for the No-U-Turn Sampler.
- max_treedepth
max value in exponents of 2 of what the binary tree size in the NUTS algorithm should have
- seed
random seed for the sampler
- include_hosp
whether or not to include hospital admissions data in the likelihood (default = 1) to include
- compute_likelihood
whether or not to include any data in the likelihood (default = 1), if set to 0 returns prior predictions
- n_draws
number of draws to save in the draws.parquet, default = 100
- n_chains
number of independent MCMC chains to run, default = 4
- iter_sampling
number of iterations to save in MCMC sampling, default = 500
- iter_warmup
number of iterations to discard in MCMC sampling, default = 250
- n_parallel_chains
number of chains to run in parallel, default = 4
- write_files
whether or not to save the outputs, default = 1 to save
- output_full_df
whether or not to have targets return the full dataframe of draws, default = FALSE to just return a dataframe of filepaths
- ...
Additional named arguments (ignored) to allow
do.call()on lists with additional entries