metajags: Compile metajags code into JAGS code

In mjskay/metabayes: Meta-Languages for Specifying Bayesian Models (JAGS, Stan) as R Code

Description

Compiles metajags code (specified as an R expression) into JAGS code.

Usage

metajags(model, data=NULL, ...)

Arguments

model	Metajags code for the JAGS model block, specified as a bare R expression.
data	Metajags code for the JAGS data block, specified as a bare R expression.
...	Metajags code for other JAGS blocks, specified as bare R expressions. The name of the argument to this function is used as the name of the JAGS block in the compiled code.

Details

metabayes integrates JAGS model specification more easily into R by allowing JAGS models to be specified as bare R code rather than as character strings. Because R and JAGS are syntactically similar, with only a few exceptions, metajags models look exactly like JAGS models. This approach has the advantage that syntax checking in R editors helps prevent simple errors without having to attempt to compile the model with JAGS, decreasing turnaround time when iterating on models.

To specify a model in metajags, simply write the code for each named JAGS block as an R expression and pass that code into this function.

Value

An object of class c("metajags", "metamodel"). Metajags models have the following functions:

code	Returns the JAGS code for this model as a character string.

Author(s)

Matthew Kay

See Also

See also code for extracting the resulting JAGS model as a character string, and metastan for the Stan equivalent of this function.

Examples

## Not run: 
library(runjags)

# FIRST, using metajags:
model = metajags(
    model = {
        #core model
        for (i in 1:n) {
            # latent variable log-linear model
            mu[i] <- b[1] + b[2]*x[i]
            y[i] ~ dnorm(mu[i], tau)
        }
        
        #priors
        b[1] ~ dnorm(0, 10)
        b[2] ~ dnorm(0, 10)
        tau ~ dgamma(0.01, 0.01)
    }
)

#(some code setting up data_list, etc) 
#...

jags_fit = run.jags(code(model), data=data_list, ...)


#SECOND, the traditional way: The above metajags approach 
#is equivalent to the following more traditional approach:

model_string = "
    model {
        #core model
        for (i in 1:n) {
            # latent variable log-linear model
            mu[i] <- b[1] + b[2]*x[i]
            y[i] ~ dnorm(mu[i], tau)
        }
        
        #priors
        b[1] ~ dnorm(0, 10)
        b[2] ~ dnorm(0, 10)
        tau ~ dgamma(0.01, 0.01)
    }
"

jags_fit = run.jags(model_string, data=data_list, ...)

## End(Not run)

mjskay/metabayes documentation built on May 23, 2019, 1:05 a.m.