sampler: Fit 'bdm' model

In cttedwards/bdm: Bayesian biomass dynamics model

Fit bdm model

Description

Execute a Bayesian model fit using rstan.

Usage

sampler(object, ...)

## S4 method for signature 'bdm'
sampler(
  object,
  data = list(),
  run = character(),
  init = "random",
  chains,
  iter,
  warmup,
  thin,
  ...
)

Arguments

object	a bdm model object
...	further arguments to sampling
data	a list object containing the model inputs
run	optional character vector to label the run
init	an initialisation list, function or character string
chains	number of MCMC chains
iter	number of iterations per chain
warmup	number of iterations to be discarded
thin	sampling interval from chains

Details

By default a Bayesian fit is executed through a call to sampling, which implements an MCMC algorithm. Default values for chains, iter, warmup and thin follow those for rstan.

The init argument can be a list, function or character string. If it is a function then it should take no arguments and return a named list of intial values for the estimated parameters. Alternatively the list can be specified directly. This behaviour matches that for sampling. If a character string is supplied it should be either 'random' or 'fixed'. If the model is the default model and init = 'fixed' then sensible starting values for r, logK and x are produced using getr, getlogK and getx. If the model is the default model and init = 'random' then sensible starting values are obtained by sampling from the priors for r, logK, x. If the model is not the default model, then the user should specify a function or list, otherwise starting values will be randomly generated by sampling(init = 'random', ...).

Value

Returns a bdm object containing posterior samples contained in object@trace.

Examples

# get some data
data(albio)
dat <- bdmData(harvest = albio$catch, index = albio$cpue, time = rownames(albio))

# initialize and fit default model
## Not run: 
mdl <- bdm()
mdl <- compiler(mdl)
mdl <- sampler(mdl, dat)

## End(Not run)

cttedwards/bdm documentation built on Oct. 11, 2022, 7:52 p.m.