adapt_PsMCMC: Pseudo-Marginal MCMC Adaptation

In JMacDonaldPhD/REpi: Pseudo-Marginal Analysis of partially observed Epidemics

Pseudo-Marginal MCMC Adaptation

Description

Adapts proposal parameters of Pseudo-Marginal MCMC scheme to give near-optimal performance.

Usage

adapt_PsMCMC(
  init,
  epiModel,
  obsFrame,
  epiSample,
  I0,
  alpha,
  logPrior,
  lambda0,
  V0,
  delta = 0.05,
  N,
  noIts
)

Arguments

init	initial value of epidemic parameters
epiModel	epidemic model
obsFrame	Generator function for observational model.
epiSample	Observed epidemic data.
I0	Initial state of the epidemic, which assumed to be known.
alpha	Observational parameters to be passed on to the log-likelihood function generated by obsFrame(X_sim)
logPrior	Functions which calculate log-density of the assumed priors of the epidemic parameters.
lambda0	Initial value of the Random Walk Metropolis (RWM) proposal scale parameter
V0	A square matrix containing the initial covariance values for RWM proposal. Shape of matrix should match the length of the init parameter.
delta	Probability that the initial proposal parameters are used as opposed to using the current adapted proposal parameters.
N	Number of particles to be used in Importance Sample Estimate of posterior. This can be optimised using adapt_IS()
noIts	Number of iterations of MCMC sampler scheme to carry out.

Value

A list of the adapted proposal parameters.

JMacDonaldPhD/REpi documentation built on Aug. 2, 2022, 2:09 p.m.