do_admb: Compile and/or run an ADMB model, collect output

In R2admb: ADMB to R interface functions

Description

Compile an ADMB model, run it, collect output

Usage

do_admb(fn, data, params,
bounds=NULL,
phase=NULL,
re = NULL,
data_type = NULL,
safe = TRUE,
profile = FALSE,
profpars = NULL,
mcmc = FALSE,
mcmc.opts = mcmc.control(),
impsamp = FALSE,
verbose=FALSE,
run.opts=run.control(),
objfunname="f",
workdir=getwd(),
admb_errors=c("stop","warn","ignore"),
extra.args)

Arguments

fn	(character) base name of a TPL function, located in the working directory
data	a list of input data variables (order must match TPL file)
params	a list of starting parameter values (order must match TPL file)
bounds	named list of 2-element vectors of lower and upper bounds for specified parameters
phase	named numeric vector of phases (not implemented yet)
re	a named list of the identities and dimensions of any random effects vectors or matrices used in the TPL file
data_type	a named vector specifying (optional) data types for parameters, in parname="storage mode" format (e.g. c(x="integer",y="numeric"))
safe	(logical) compile in safe mode?
profile	(logical) generate likelihood profiles? (untested!)
profpars	(character) vector of names of parameters to profile
mcmc	(logical) run MCMC around best fit?
mcmc.opts	options for MCMC (see mcmc.control for details)
impsamp	(logical) run importance sampling?
verbose	(logical) print details
run.opts	options for ADMB run (see run.control for details)
objfunname	(character) name for objective function in TPL file (only relevant if checkparam is set to "write")
workdir	temporary working directory (dat/pin/tpl files will be copied)
admb_errors	how to treat ADMB errors (in either compilation or run): use at your own risk!
extra.args	(character) extra argument string to pass to admb

Details

do_admb will attempt to do everything required to start from the model definition (TPL file) specified by fn, the data list, and the list of input parameters, compile and run (i.e. minimize the objective function of) the model in AD Model Builder, and read the results back into an object of class admb in R. If checkparam or checkdata are set to "write", it will attempt to construct a DATA section, and construct or (augment an existing) PARAMETER section (which may contain definitions of non-input parameters to be used in the model). It copies the input TPL file to a backup (.bak); on finishing, it restores the original TPL file and leaves the auto-generated TPL file in a file called [fn]_gen.tpl.

Value

An object of class admb.

Note

1. Mixed-case file names are ignored by ADMB; this function makes a temporary copy with the file name translated to lower case. 2. Parameter names containing periods/full stops will not work, because this violates C syntax (currently not checked). 3. There are many, many, implicit restrictions and assumptions: for example, all vectors and matrices are assumed to be indexed starting from 1.

Author(s)

Ben Bolker

Examples

## Not run: 
setup_admb()
file.copy(system.file("tplfiles","ReedfrogSizepred0.tpl",package="R2admb"),"tadpole.tpl")
 tadpoledat <-
  data.frame(TBL = rep(c(9,12,21,25,37),each=3),
             Kill = c(0,2,1,3,4,5,0,0,0,0,1,0,0,0,0L),
             nexposed=rep(10,15))
m1 <- do_admb("tadpole",
              data=c(list(nobs=15),tadpoledat),
              params=list(c=0.45,d=13,g=1),
              bounds=list(c=c(0,1),d=c(0,50),g=c(-1,25)),
              run.opts=run.control(checkparam="write",
                checkdata="write",clean="all"))
unlink("tadpole.tpl")

## End(Not run)

R2admb documentation built on May 2, 2019, 5:19 p.m.