init_settings: Initialize a bookkeeping list of MCMC settings.

In fintzij/BDAepimodel: Fit Stochastic Epidemic Models in R via Bayesian Data Augmentation

Description

Initialize a bookkeeping list of MCMC settings.

Usage

init_settings(epimodel, niter, configs_to_redraw,
  config_replacement = FALSE, preferential_sampling = NULL,
  init_popsize = NULL, compartment_dist = NULL,
  save_params_every = 1, save_configs_every = niter%/%100, kernel,
  post_init_params = NULL, cov_mtx = NULL,
  to_estimation_scale = NULL, from_estimation_scale = NULL,
  analytic_eigen = NULL, ecctmc_method = "mr", seed = NULL,
  time_limit = NULL)

Arguments

epimodel	epimodel object
niter	number of iterations/parameter updates for which to run the sampler.
configs_to_redraw	number of subject level trajectories to sample between parameter updates.
preferential_sampling	vector with two elements, given in the following order. First the size of the group to be preferentially sampled. Second, the probability that a subject is sampled from the preferential group. The IDs of subjects to be preferentially sampled will be selected uniformly at random from among subjects with non-constant paths when the collection of subject paths is initialized. If the size of the preferentially sampled group is larger than the number of subjects with non-constant paths, additional subject IDs will be added by selecting subejct IDs uniformly at random from among the other subjects.
init_popsize	population size with which to initialize the epidemic.
compartment_dist	distribution according to which to place the remainder of the subjects. required if init_popsize is not equal to the true population size. Supplied as a named vector of probabilities.
save_params_every	thin parameter updates by saving every k-th draw, defaults to 1.
save_configs_every	thin configurations by saving the current configuration after every k-th parameter update.
kernel	transition kernel for parameters, provided as a list of transition functions. Each function should take as an argument only the epimodel list. The kernels should be self-contained, in particular making any necessary transformations of parameters and calculating prior probabilities as needed. Each kernel function should include a call to update_params at the end.
post_init_params	a vector of initial parameter values to be used after an acceptable initial configuration has been simulated.
cov_mtx	covariance matrix to be used in the transition kernels.
to_estimation_scale	list of functions for transforming model parameters to the scale on which new parameters should be proposed. List element names should correspond to exactly to the parameter names given in params.
from_estimation_scale	list of functions for transforming model parameters from the a scale on which new parameters are proposed to the scale used to evaluate the process likelihood. List element names should correspond to exactly to the parameter names given in params.
analytic_eigen	optional. If NULL, eigenvalues, eigenvectors, and inverses of the eigenvector matrices of each rate transition rate matrix are computed numerically. If one of "SIR", or "SEIR", the eigen decompositions will be computed analytically. It is up to the user to ensure that the rate matrices are structured in the appropriate form (see vignettes for examples).
ecctmc_method	Method for sampling the exact times of state transition. Defaults to "mr" for modified rejection sampling. May also be specified as "unif" for uniformization.
seed	optional seed value. A random seed is generated and saved if none is supplied.
time_limit	optional maximum time budget for the MCMC, checked every time the parameters are saved. Should be provided as a diffitme object with units in hours.

Value

bookkeeping list

fintzij/BDAepimodel documentation built on Sept. 20, 2020, 1:44 p.m.