Control parameters for tinyVAST

Usage

tinyVASTcontrol(
  nlminb_loops = 1,
  newton_loops = 0,
  eval.max = 1000,
  iter.max = 1000,
  getsd = TRUE,
  silent = getOption("tinyVAST.silent", TRUE),
  trace = getOption("tinyVAST.trace", 0),
  verbose = getOption("tinyVAST.verbose", FALSE),
  profile = c(),
  tmb_par = NULL,
  tmb_map = NULL,
  gmrf_parameterization = c("separable", "projection"),
  reml = FALSE,
  getJointPrecision = FALSE,
  calculate_deviance_explained = TRUE,
  run_model = TRUE,
  suppress_nlminb_warnings = TRUE,
  suppress_user_warnings = FALSE,
  get_rsr = FALSE,
  extra_reporting = FALSE,
  use_anisotropy = FALSE,
  sar_adjacency = "queen",
  barrier_stiffness = 0.01
)

Arguments

nlminb_loops: Integer number of times to call stats::nlminb().
newton_loops: Integer number of Newton steps to do after running stats::nlminb().
eval.max: Maximum number of evaluations of the objective function allowed. Passed to control in stats::nlminb().
iter.max: Maximum number of iterations allowed. Passed to control in stats::nlminb().
getsd: Boolean indicating whether to call TMB::sdreport()
silent: Disable terminal output for inner optimizer?
trace: Parameter values are printed every trace iteration for the outer optimizer. Passed to control in stats::nlminb().
verbose: Output additional messages about model steps during fitting?
profile: Character-vector passed to TMB::MakeADFun and see description there. Fixed effects that are highly correlated with random effects can often be estimated faster (i.e., with fewer iterations) by adding them to profile. The most common use-case is profile = c("alpha_j","alpha2_j"). However, doing so will have a small impact on model estimates and predictions.
tmb_par: list of parameters for starting values, with shape identical to tinyVAST(...)$internal$parlist
tmb_map: input passed to TMB::MakeADFun as argument map, over-writing the version tinyVAST(...)$tmb_inputs$tmb_map and allowing detailed control over estimated parameters (advanced feature)
gmrf_parameterization: Parameterization to use for the Gaussian Markov random field, where the default separable constructs a full-rank and separable precision matrix, and the alternative projection constructs a full-rank and IID precision for variables over time, and then projects this using the inverse-cholesky of the precision, where this projection allows for rank-deficient covariance.
reml: Logical: use REML (restricted maximum likelihood) estimation rather than maximum likelihood? Internally, this adds the fixed effects to the list of random effects to integrate over.
getJointPrecision: whether to get the joint precision matrix. Passed to sdreport.
calculate_deviance_explained: whether to calculate proportion of deviance explained. See deviance_explained()
run_model: whether to run the model of export TMB objects prior to compilation (useful for debugging)
suppress_nlminb_warnings: whether to suppress uniformative warnings from nlminb arising when a function evaluation is NA, which are then replaced with Inf and avoided during estimation
suppress_user_warnings: whether to suppress warnings from package author regarding dangerous or non-standard options
get_rsr: Experimental option, whether to report restricted spatial regression (RSR) adjusted estimator for covariate responses
extra_reporting: Whether to report a much larger set of quantities via obj$env$report()
use_anisotropy: Whether to estimate two parameters representing geometric anisotropy
sar_adjacency: Whether to use queen or rook adjacency when defining a Simultaneous Autoregressive spatial precision from a sfc_GEOMETRY (default is queen)
barrier_stiffness: The ratio of local stiffness (the scale of diffusion rate and resulting decorrelation distance) for barriers relative to normal areas in the SPDE method when using add_mesh_covariates. The default barrier_stiffness = 0.01 is the value from Bakka et al. 2019.

Value

An object (list) of class tinyVASTcontrol, containing either default or updated values supplied by the user for model settings

References

Bakka, H., Vanhatalo, J., Illian, J., Simpson, D., Rue, H. (2019). Non-stationary Gaussian models with physical barriers. Spatial Statistics, 29, 268-288. doi:10.1016/j.spasta.2019.01.002