Simulates new datasets from the fitted model. For each
posterior draw of the model parameters, generates a complete replicated
dataset. These replications can be compared to the observed data using
plot_ppc_ecdf() or plot_ppc_density() to check whether the model
captures the key features of the data (a "posterior predictive check").
Arguments
- fit
A dbn model fit object (output from
dbn())- ndraws
Number of replicated datasets to generate (default: 100)
- seed
Random seed for reproducibility
- draws
Specific posterior draw indices to use (overrides
ndraws)
Value
A list of ndraws replicated data arrays, each with the same
dimensions as the original data. Has class "dbn_ppd".
Examples
# \donttest{
sim <- simulate_dynamic_dbn(n = 6, time = 5, seed = 1)
fit <- dbn(sim$Y, model = "dynamic", nscan = 200, burn = 100, verbose = FALSE)
ppd <- posterior_predict_dbn(fit, ndraws = 5)
# }