Description Usage Arguments Value Author(s) Examples
This function wraps all the necessary functions in DAbayes package to run the MCMC Suite. By default, this will execute the adaptive MCM algorithm in single core (sequentially), and it can be executed in parallel by changing the number of cores to the number of cores available.
1 2 3 | DAbayesSuite(ensemble, model_runs, control_runs, r_max = NULL,
theta_intVal = NULL, prior = NULL, AD_proposal = NULL, niter = 20000,
start_adapting = 50, numCore = 1)
|
ensemble |
an n by N matrix of ensembles of measured variable, say temperature increase |
model_runs |
a list of elements, each of which is model-output of the measured variable (temperature increase) under a specific forcing scenario |
control_runs |
an n by L0 matrix, where n is the number of grid cells, and L0 is the number of control runs |
r_max |
an integer representing the maximum number of empirical orthogonal functions |
theta_intVal |
a list of 3 elements containing initial values for parameters in the MCMC algorithm |
prior |
a list of 6 elements in prior distributions for corresponding parameters |
AD_proposal |
a list of 6 elements containing quantities to adjust mean and covariance in the proposal distribution |
niter |
an integer containing the total number of MCMC interations |
start_adapting |
an integer specifiying when to adapt proposal in the MCMC algorithm |
numCore |
an integer specifying how many cores are used in parallel computing |
a list of r elements, where r is the number of EOFs specified, and each element is again 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
Pulong Ma <mpulong@gmail.com>
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 | ## Not run:
## Run Gibbs sampling with Metropolis-Hastings algorithm in the
# Bayesian statistical model
data(ensemble_temperature)
data(GCM_runs)
data(GCM_control_run)
Res <- DAbayesSuite(ensemble=ensemble_temperature, model_runs=GCM_runs,
control_runs=GCM_control_run, r_max=2, theta_intVal=NULL, prior=NULL,
AD_proposal=NULL, niter=20000, start_adapting=100, numCore=1)
## diagnostic plots for posterior samples, and more sophisticated
# diagnostic plots can be obtained with coda package
# take the first element in Res for example
# plot only first 2 beta's
plot(Res[[1]]$beta[1, ], xlab="Iteration", ylab="beta_1", type="l")
plot(Res[[1]]$beta[2, ], xlab="Iteration", ylab="beta_2", type="l")
plot(Res[[1]]$logsigma, xlab="Iteration", ylab="logsigma", type="l")
# plot only first 3 lambda's
plot(Res[[1]]$lambda[1, ], xlab="Iteration", ylab="lambda1", type="l")
plot(Res[[1]]$lambda[2, ], xlab="Iteration", ylab="lambda2", type="l")
plot(Res[[1]]$lambda[3, ], xlab="Iteration", ylab="lambda3", type="l")
## save plots
burnin <- 5000
plotMCMC(Res, N, n, dir=getwd(), burnin)
## End(Not run)
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.