run_admb_mcmc: Run an MCMC using an ADMB model, return (1) the posterior...

Description Usage Arguments Value

Description

Run an MCMC using an ADMB model, return (1) the posterior draws, MLE fits and covariance/correlation matrices, and some MCMC convergence diagnostics using CODA.

Run an MCMC using an ADMB model, return (1) the posterior draws, MLE fits and covariance/correlation matrices, and some MCMC convergence diagnostics using CODA.

Usage

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
run_admb_mcmc(model.path, model.name, iter, mcsave, burn.in, cov.user = NULL,
  init.pin = NULL, se.scale = NULL, mcscale = FALSE, mcseed = NULL,
  mcrb = NULL, mcdiag = FALSE, mcprobe = NULL, verbose = TRUE,
  extra.args = NULL, hybrid = FALSE, hyeps = NULL, hynstep = NULL,
  mceval = FALSE, estimate = FALSE)

run_admb_mcmc(model.path, model.name, iter, mcsave, burn.in, cov.user = NULL,
  init.pin = NULL, se.scale = NULL, mcscale = FALSE, mcseed = NULL,
  mcrb = NULL, mcdiag = FALSE, mcprobe = NULL, verbose = TRUE,
  extra.args = NULL, hybrid = FALSE, hyeps = NULL, hynstep = NULL,
  mceval = FALSE, estimate = FALSE)

Arguments

model.path

(Character) A path to the folder containing the model. NULL indicates the current folder.

iter

(Integer) The number of draws after thinning and burn in.

mcsave

(Integer) Controls thinning of samples. Save every mcsave value, such that 1 corresponds to keeping all draws, and 100 saving every 100th draw.

burn.in

(Integer) How many samples to discard from the beginning of the chain, *after* thining. The burn in period (i.e., the first burn.in*mcsave draws) should be at least large enough to cover dynamic scaling.

cov.user

(Numeric matrix) A manually defined covariance matrix (in bounded space) to use in the Metropolis-Hastings algorithm.

init.pin

(Numeric vector) A vector of initial values, which are written to file and used in the model via the -mcpin option.

se.scale

(Numeric) A value which scales all of the variances from the MLE fit. A value of 1 indicates to use the estimated variances.

mcscale

(Logical) Whether to use the mcscale option, which dynamically scales the covariance matrix for efficient acceptance ratios.

mcseed

(Integer) Which seed (integer value) to pass ADMB. Used for reproducibility.

mcrb

(Integer) Which value to use in the rescale bounded algorithm. Must be an integer from 1-9. The default NULL value disables this feature. See the vignette for more information on this algorithm and how to best use it.

mcdiag

(Logical) Whether to use the mcdiag feature. This uses an identity matrix for the covariance matrix.

verbose

(Logical) Whether to print ADMB warnings and other information. Useful for testing and troubleshooting.

extra.args

(Character) A string which is passed to ADMB at runtime. Useful for passing additional arguments to the model executable.

hyeps

(Numeric) The size of the leapfrog jump in the hybrid method, with smaller values leading to smaller but more accurate jumps. Must be a positive value.

hynstep

(Integer) The approximate number of steps used in the leapfrog step of the hybrid algorithm. Steps are randomly generated for each MCMC iteration, centered around hynstep.

mode.name

(Character) The name of the model executable. A character string, without '.exe'.

thin

(Integer) Controls thinning of samples. Save every thin value, such that 1 corresponds to keeping all draws, and 100 saving every 100th draw.

warmup

(Integer) How many samples to discard from the beginning of the chain, *after* thining. The burn in period (i.e., the first warmup*thin draws) should be at least large enough to cover dynamic scaling.

init

(Numeric vector) A vector of initial values, which are written to file and used in the model via the -mcpin option.

eps

(Numeric) The size of the leapfrog jump in the hybrid method, with smaller values leading to smaller but more accurate jumps. Must be a positive value.

model.path

(Character) A path to the folder containing the model. NULL indicates the current folder.

mode.name

(Character) The name of the model executable. A character string, without '.exe'.

iter

(Integer) The number of draws after thinning and burn in.

cov.user

(Numeric matrix) A manually defined covariance matrix (in bounded space) to use in the Metropolis-Hastings algorithm.

mcseed

(Integer) Which seed (integer value) to pass ADMB. Used for reproducibility.

mcdiag

(Logical) Whether to use the mcdiag feature. This uses an identity matrix for the covariance matrix. #' @param mcprobe Which value to use in the probing algorithm. The default NULL value disables this feature. See the vignette for more information on this algorithm and how to best use it.

hyrbid

(Logical) Whether to use the Hamiltonial (hybrid) algorithm. Default is FALSE.

verbose

(Logical) Whether to print ADMB warnings and other information. Useful for testing and troubleshooting.

extra.args

(Character) A string which is passed to ADMB at runtime. Useful for passing additional arguments to the model executable.

Value

Returns a list containing (1) the posterior draws, (2) and object of class 'admb', read in using the results read in using read_admb, and (3) some MCMC convergence diagnostics using CODA.

Returns a list containing (1) the posterior draws, (2) and object of class 'admb', read in using the results read in using read_admb, and (3) some MCMC convergence diagnostics using CODA.


colemonnahan/admbtools documentation built on May 13, 2019, 8:50 p.m.