smesir: smesir: Fitting Semiparametric Mixed-Effects SIR Models

In davidbuch/smesir: Modeling regional incidence data to understand transmission rates

smesir: Fitting Semiparametric Mixed-Effects SIR Models

Description

This function fits an SIR model with local, time-varying transmission rates to epidemic incidence data (deaths) observed in one or more regions. The transmission rate in each region is modeled as a linear function of covariates plus a smooth temporal random effect drawn from a Gaussian Process distribution. The intercepts, coefficients, and even the temporal random effects from the various regions are assumed to be independent draws from global distributions centered at "global" intercept, coefficient, and temporal random effect values.

Usage

smesir(
  formula,
  data,
  epi_params,
  region_names = NULL,
  prior = NULL,
  chains = 4,
  iter = 50000,
  warmup = 0,
  thin = max(floor((iter - warmup)/1000), 1),
  min_adaptation_cycles = 5,
  min_samps_per_cycle = NULL,
  tempering_ratio = 0.2,
  quiet = TRUE,
  sr_style = NULL,
  seed = NULL
)

Arguments

formula	Object of class "formula"
data	A named list containing the data with which the model will be fit. The list should include an entry for each term in the accompanying formula argument.
epi_params	Epidemiologic parameters which must be specified by the user (typically obtained from side-information): region_populations - Vector of populations of the regions studied, listed in the same order in which they are indexed in the data outbreak_times - Vector of indices of the time interval at which the first cases are reported in each region mean_removal_time - Average amount of time (in number or fractions of time intervals) that an individual remains infectious psi - Vector of probabilities whose element i is the probability that the response event (case detection, death) occurs an infected individual i - 1 time intervals after their infection
region_names	Vector of names of the regions studied, listed in the same order in which they are indexed in the data
prior	(Optional - reasonable default values are specified internally) A length 4 named list containing: ell - lengthscale of the squared exponential kernal for the temporal random effect V0 - 3 variance hyperparameters for gaussian priors on the intercepts, coefficients, and temporal random effects; IGSR - 3 pairs of shape and rate hyperparameters for inverse-gamma priors; expected_initial_infected - the expected size of the infected population that appears at the beginning of the outbreak, used in an exponential prior;

davidbuch/smesir documentation built on Oct. 31, 2022, 1:14 p.m.