nimbleModel: Create a NIMBLE model from BUGS code

In nimble: MCMC, Particle Filtering, and Programmable Hierarchical Modeling

Create a NIMBLE model from BUGS code

Description

processes BUGS model code and optional constants, data, and initial values. Returns a NIMBLE model (see modelBaseClass) or model definition.

Usage

nimbleModel(
  code,
  constants = list(),
  data = list(),
  inits = list(),
  dimensions = list(),
  returnDef = FALSE,
  where = globalenv(),
  debug = FALSE,
  check = getNimbleOption("checkModel"),
  calculate = TRUE,
  name = NULL,
  buildDerivs = getNimbleOption("buildModelDerivs"),
  userEnv = parent.frame()
)

Arguments

code	code for the model in the form returned by nimbleCode or (equivalently) quote
constants	named list of constants in the model. Constants cannot be subsequently modified. For compatibility with JAGS and BUGS, one can include data values with constants and nimbleModel will automatically distinguish them based on what appears on the left-hand side of expressions in code.
data	named list of values for the data nodes. Data values can be subsequently modified. Providing this argument also flags nodes as having data for purposes of algorithms that inspect model structure. Values that are NA will not be flagged as data.
inits	named list of starting values for model variables. Unlike JAGS, should only be a single list, not a list of lists.
dimensions	named list of dimensions for variables. Only needed for variables used with empty indices in model code that are not provided in constants or data.
returnDef	logical indicating whether the model should be returned (FALSE) or just the model definition (TRUE).
where	argument passed to setRefClass, indicating the environment in which the reference class definitions generated for the model and its modelValues should be created. This is needed for managing package namespace issues during package loading and does not normally need to be provided by a user.
debug	logical indicating whether to put the user in a browser for debugging. Intended for developer use.
check	logical indicating whether to check the model object for missing or invalid values. Default is given by the NIMBLE option 'checkModel'. See nimbleOptions for details.
calculate	logical indicating whether to run calculate on the model after building it; this will calculate all deterministic nodes and logProbability values given the current state of all nodes. Default is TRUE. For large models, one might want to disable this, but note that deterministic nodes, including nodes introduced into the model by NIMBLE, may be NA.
name	optional character vector giving a name of the model for internal use. If omitted, a name will be provided.
buildDerivs	logical indicating whether to build derivative capabilities for the model.
userEnv	environment in which if-then-else statements in BUGS code will be evaluated if needed information not found in constants; intended primarily for internal use only

Details

See the User Manual or help(modelBaseClass) for information about manipulating NIMBLE models created by nimbleModel, including methods that operate on models, such as getDependencies.

The user may need to provide dimensions for certain variables as in some cases NIMBLE cannot automatically determine the dimensions and sizes of variables. See the User Manual for more information.

As noted above, one may lump together constants and data (as part of the constants argument (unlike R interfaces to JAGS and BUGS where they are provided as the data argument). One may not provide lumped constants and data as the data argument.

For variables that are a mixture of data nodes and non-data nodes, any values passed in via inits for components of the variable that are data will be ignored. All data values should be passed in through data (or constants as just discussed).

Author(s)

NIMBLE development team

See Also

readBUGSmodel for creating models from BUGS-format model files

Examples

code <- nimbleCode({
    x ~ dnorm(mu, sd = 1)
    mu ~ dnorm(0, sd = prior_sd)
})
constants = list(prior_sd = 1)
data = list(x = 4)
Rmodel <- nimbleModel(code, constants = constants, data = data)

nimble documentation built on July 9, 2023, 5:24 p.m.