calculateMarginalLogLikelihood {AnaCoDa} | R Documentation |
initializes the model object.
calculateMarginalLogLikelihood(
parameter,
mcmc,
mixture,
n.samples,
divisor,
warnings = TRUE
)
parameter |
An object created with |
mcmc |
An object created with |
mixture |
determines for which mixture the marginal log-likelihood should be calculated |
n.samples |
How many samples should be used for the calculation |
divisor |
A value > 1 in order to scale down the tails of the importance distribution |
warnings |
Print warnings such as when the variance of a parameter is 0, which might occur when parameter is fixed |
calculateMarginalLogLikelihood Calculate marginal log-likelihood for calculation of the Bayes factor using a generalized harmonix mean estimator of the marginal likelihood. See Gronau et al. (2017) for details
This function returns the model object created.
## Not run:
# Calculate the log-marginal likelihood
parameter <- loadParameterObject("parameter.Rda")
mcmc <- loadMCMCObject("mcmc.Rda")
calculate_marginal_likelihood(parameter, mcmc, mixture = 1,
samples = 500, scaling = 1.5)
# Calculate the Bayes factor for two models
parameter1 <- loadParameterObject("parameter1.Rda")
parameter2 <- loadParameterObject("parameter2.Rda")
mcmc1 <- loadMCMCObject("mcmc1.Rda")
mcmc2 <- loadMCMCObject("mcmc2.Rda")
mll1 <- calculate_marginal_likelihood(parameter1, mcmc1, mixture = 1,
samples = 500, scaling = 1.5)
mll2 <- calculate_marginal_likelihood(parameter2, mcmc2, mixture = 1,
samples = 500, scaling = 1.5)
cat("Bayes factor: ", mll1 - mll2, "\n")
## End(Not run)