Benchmark results

Evaluate outputs from models implemented using a (1) rjMCMC and (2) Gibbs Variable Selection approach. Returns comparative plots of posterior model rankings and posterior parameter estimates.

compare_models(
  rj.dat = NULL,
  gvs.fixed = NULL,
  gvs.random = NULL,
  by.model = FALSE,
  kernel.adj = 2,
  viridis.col = FALSE,
  density = TRUE,
  prob = TRUE
)

Arguments

rj.dat	Input rjMCMC object, as returned by trace_rjMCMC.
gvs.fixed	Input Gibbs object, as returned by gibbs using a fixed effect implementation of the dose-response model.
gvs.random	Input Gibbs object, as returned by gibbs using a random effect implementation of the dose-response model.
by.model	Logical. If TRUE, the functions subsets posterior estimates by candidate model.
kernel.adj	Bandwidth adjustment. The bandwidth used to create density plots is given by kernel.adj*bw.
viridis.col	Logical. Whether to use a viridis colour scheme.
density	Logical. If TRUE, compares density plots for each model parameter.
prob	Logical. If TRUE, compares posterior rankings for candidate models.

See also

simulate_data example_brs summary.rjdata

Author

Phil J. Bouchet

Examples

if (FALSE) {
library(espresso)

# Simulate data for two species
mydat <- simulate_data(n.species = 2, 
                       n.whales = 16, 
                       max.trials = 3, 
                       covariates = list(exposed = c(0, 5), range = 0.5),
                       mu = c(101, 158), 
                       phi = 20, 
                       sigma = 20, 
                       Rc = c(210, 211), 
                       seed = 58697)
summary(mydat)
     
# Model selection by GVS                        
gvs <- gibbs(dat = mydat, 
             random.effects = FALSE, 
             include.covariates = FALSE, 
             mcmc.n = 1000, 
             burnin = 500)

# Run the reversible jump MCMC
rj <- run_rjMCMC(dat = mydat.config,
                 n.chains = 2,
                 n.burn = 100,
                 n.iter = 100,
                 do.update = FALSE)
                 
# Burn and thin
rj.trace <- trace_rjMCMC(rj.dat = rj)

# Compare outputs
compare_models(rj.dat = rj.trace, gvs.fixed = gvs)
}