acceptProposal: Determine if a Metropolis–Hastings step should be accepted
In cbdAmgen/bcgp0a: Bayesian Composite Gaussian Processes

Description Usage Arguments Details Value Examples

acceptProposal is a utility function to determine if a proposal should be accepted in a Metropolis or Metropolis-Hastings step. This is shamelessly stolen from AcceptProposal in the overture package.

1	acceptProposal(logCurr, logProp, logCurrToProp = 0, logPropToCurr = 0)

`logCurr`	log density of the target at the current value, log(P(x))
`logProp`	log density of the target at the proposed value, log(P(x'))
`logCurrToProp`	log of transition distribution from current value to proposed value, log(g(x'\|x))
`logPropToCurr`	log of transition distribution from proposed value to current value, log(g(x\|x'))

The function uses the Metropolis choice for a Metropolis/Metropolis-Hastings sampler, which accepts a proposed value x' with probability

A(x', x) = min(1, P(x')/P(x) g(x|x')/g(x'|x))

where P(x) is the target distribution and g(x'|x) is the proposal distribution.

TRUE/FALSE for whether the proposal should be accepted or rejected, respectively

curr <- rnorm(1, 0, 1)
prop <- curr + runif(1, -0.5, 0.5)
acceptProposal(logCurr = dnorm(curr, 0, 1, log = TRUE),
               logProp = dnorm(prop, 0, 1, log = TRUE))