Fit Bayesian model using MCMC

Source: R/models.R

This is a general wrapper function for fitting a model to data using the JAGS MCMC algorithm. The function uses code adapted from the rjags package. Options include calculating the Deviance Information Criterion (DIC) and running sampling chains in parallel.

fit_jags(
  jags_data,
  jags_model,
  params,
  n_chain = 2,
  n_burn = 1000,
  n_samp = 1000,
  n_thin = 1,
  DIC = FALSE,
  parallel = FALSE
)

Arguments

jags_data

a list of data objects
jags_model

character string giving the model specification in JAGS/BUGS syntax
params

vector of parameters to monitor
n_chain

number of MCMC sampling chains
n_burn

number of iterations to discard before sampling of chains begins (burn in)
n_samp

number of iterations to sample each chain
n_thin

interval to thin samples
DIC

logical indicating whether or not to calculate the Deviance Information Criterion (DIC)
parallel

logical indicating whether or not to run MCMC chains in parallel or sequentially (default = FALSE). An even number of chains is recommended when running in parallel.

Value

an mcmc.list object

Details

If JAGS library not already installed, install here: https://sourceforge.net/projects/mcmc-jags/files/JAGS/3.x/

See also

Other model: check(), compare(), fit_prob_travel(), mobility(), predict(), residuals(), summary()

Author

John Giles

Examples

N <- 100
lambda <- 10
y <- rpois(n=N, lambda=lambda)

jags_data <- list(N=N, y=y)

jags_model <- "
model {

  for(i in 1:N){

    y[i] ~ dpois(lambda)

  }

  lambda ~ dgamma(1, 0.01)

}"

params <- c('lambda')

mod <- fit_jags(jags_data=jags_data,
                jags_model=jags_model,
                params=params)

#> Compiling model graph
#>    Resolving undeclared variables
#>    Allocating nodes
#> Graph information:
#>    Observed stochastic nodes: 100
#>    Unobserved stochastic nodes: 1
#>    Total graph size: 104
#> 
#> Initializing model
#>