parDA_GIBBS: Gibbs sampling with Metropolis-Hastings algorithm for the...

In mapn/DAbayes: Bayesian Statistical Model for Climate-Change Detection and Attribution

Description

This function implements Gibbs sampling with Metropolis-Hasting algorithm to sample from posterior distributions for the proposed Bayesian statistical model. This function is used in parallel computing of MCMC algorithm

Usage

parDA_GIBBS(outW, theta, prior_theta, ADtheta, N, Lj, niter = 20000,
  start_adapting = 500)

Arguments

outW	a 7 by 1 list containing precomputed quantities associated with W from the output of function computeW(...)
theta	a list of 3 elements contained in parameters in the MCMC algorithm
prior_theta	a list of 2 elements, each of whcih is the quantity in the prior distribution of lambda
ADtheta	a list of 6 elements, each of which is the parameter in the corresponding proposal distribution
N	an integer representing the number of ensemble members
Lj	an m by 1 vector containing the number of runs for each forcing scenario
niter	an integer specifying the total number of MCMC iterations
start_adapting	an integer specifiying when to adapt proposal in the MCMC algorithm

Value

a list of 6 elements containing posterior quantities of parameters, log-likelihood, chisq statistics, and prior:

beta: a matrix holds the posterior samples for the parameter beta with each row corresponding to each beta

logsigma: a vector holds the posterior samples for the parameter log of sigma

lambda: a vector holds the posterior samples for the parameter lambda

loglik: a vector holds the log-likelihood evaluated with updated parameters

chisq: a vector holds chisquare statistics for residual consistency test

prior: a vector holds the prior density evaluated with updated parameters

Author(s)

Pulong Ma <[e-mail]>

See Also

DA_GIBBS

mapn/DAbayes documentation built on May 21, 2019, 11:26 a.m.