fit.vast: Fits a VAST model, compatible with VAST v8_3_0 through...

In PacificCommunity/ofp-sam-vast-utils: Utility functions for applying VAST to analysis of WCPFC tuna fisheries data.

Description

Fits a VAST model, compatible with VAST v8_3_0 through 16/12/2019 Some of the descriptions come from JT VAST package descriptions

Usage

fit.vast(
  Data_Geostat,
  RunDir,
  SaveDir,
  save.output = FALSE,
  Q_ik = NULL,
  vf.re = FALSE,
  FieldConfig = c(Omega1 = 1, Epsilon1 = 1, Omega2 = 1, Epsilon2 = 1),
  RhoConfig = c(Beta1 = 0, Beta2 = 0, Epsilon1 = 0, Epsilon2 = 0),
  ObsModel_ez = c(1, 3),
  fine_scale = TRUE,
  input.grid.res = 1,
  crop.extrap.by.data = TRUE,
  knot_method = "grid",
  n_x = 100,
  Version = "VAST_v8_3_0",
  Method = "Mesh",
  ADREPORT = TRUE,
  normalize_idx = FALSE,
  Xconfig_zcp = NULL,
  slim.output = FALSE,
  strata.sp,
  enviro,
  n_xLOC,
  newton_steps = 3,
  n.boot = 250
)

Arguments

Data_Geostat	A data-frame of i rows containing the following columns: Response_variable, Year, Lon, Lat, Spp, AreaSwept_km2, Vessel
RunDir	Path to the directory where the .cpp VAST source code is stored or compiled
SaveDir	Path to the directory where the outputs will be saved
save.output	TRUE or FALSE
Q_ik	Matrix or i rows and k covariates impacting catchability. Can be created using stats::model.matrix
vf.re	TRUE or FALSE switch indicating if vessel random effects are to be estimated. If so then the Vessel column in Data_Geostat is used
FieldConfig	Controls the number of factors estimated with the spatial and spatiotemporal random fields. default setting = c(Omega1 = 1, Epsilon1 = 1, Omega2 = 1, Epsilon2 = 1)
RhoConfig	Controls the temporal structure of the annual intercepts and the spatiotemporal random field. default setting = c(Beta1 = 0, Beta2 = 0, Epsilon1 = 0, Epsilon2 = 0)
ObsModel_ez	Controls the error structure. default setting = c(1,3) ObsModel_ez[e,1]=0 Normal ObsModel_ez[e,1]=1 Lognormal ObsModel_ez[e,1]=2 Gamma ObsModel_ez[e,1]=3 Inverse-Gaussian ObsModel_ez[e,1]=5 Negative binomial ObsModel_ez[e,1]=6 Conway-Maxwell-Poisson (likely to be very slow) ObsModel_ez[e,1]=7 Poisson (more numerically stable than negative-binomial) ObsModel_ez[e,1]=8 Compound-Poisson-Gamma, where the expected number of individuals is the 1st-component, the expected biomass per individual is the 2nd-component, and SigmaM is the variance in positive catches (likely to be very slow) ObsModel_ez[e,1]=9 Binned-Poisson (for use with REEF data, where 0=0 individual; 1=1 individual; 2=2:10 individuals; 3=>10 individuals) ObsModel_ez[e,1]=10 Tweedie distribution, where epected biomass (lambda) is the product of 1st-component and 2nd-component, variance scalar (phi) is the 1st component, and logis-SigmaM is the power ObsModel_ez[e,1]=11 Zero-inflated Poisson with additional normally-distributed variation overdispersion in the log-intensity of the Poisson distribution ObsModel_ez[e,1]=12 Poisson distribution (not zero-inflated) with log-intensity from the 1st linear predictor, to be used in combination with the Poisson-link delta model for combining multiple data types ObsModel_ez[e,1]=13 Bernoilli distribution using complementary log-log (cloglog) link from the 1st linear predictor, to be used in combination with the Poisson-link delta model for combining multiple data types ObsModel_ez[e,1]=14 Similar to 12, but also including lognormal overdispersion ObsModel_ez[e,2]=0 Conventional delta-model using logit-link for encounter probability and log-link for positive catch rates ObsModel_ez[e,2]=1 Alternative "Poisson-link delta-model" using log-link for numbers-density and log-link for biomass per number ObsModel_ez[e,2]=2 Link function for Tweedie distribution, necessary for ObsModel_ez[e,1]=8 or ObsModel_ez[e,1]=10 ObsModel_ez[e,2]=3 Conventional delta-model, but fixing encounter probability=1 for any year where all samples encounter the species ObsModel_ez[e,2]=4 Poisson-link delta-model, but fixing encounter probability=1 for any year where all samples encounter the species and encounter probability=0 for any year where no samples encounter the species
fine_scale	TRUE or FALSE. Better maps and slightly better index fit when TRUE but is slower.
input.grid.res	Resolution of extrapolation grid in kilometers.
crop.extrap.by.data	TRUE or FALSE: If TRUE then the extrapolation grid is cropped by the smooth hull surrounding the data
knot_method	knot_method whether to determine location of GMRF vertices based on the location of samples knot_method=`samples` or extrapolation-grid cells within the specified strata knot_method='grid'
n_x	Number of knots
Version	Version of VAST to use. Compatible with version "VAST_v8_3_0"
Method	Method to use for defining spatial field. default setting = "Mesh"
ADREPORT	TRUE or FALSE. Calculate the SD for the params and index?
normalize_idx	TRUE or FALSE. Normalize the index (and the SE) by dividing by the mean of the index
Xconfig_zcp	OPTIONAL, 3D array of settings for each dynamic density covariate, where the first dimension corresponds to 1st or 2nd linear predictors, second dimension corresponds to model category, and third dimension corresponds to each density covariate Xconfig_zcp[z,c,p]=0 X_itp[,,p] has no effect on linear predictor z for category c Xconfig_zcp[z,c,p]=1 X_itp[,,p] has a linear effect on linear predictor z for category c Xconfig_zcp[z,c,p]=2 X_itp[,,p] has a spatially varying, zero-centered linear effect on linear predictor z for category c Xconfig_zcp[z,c,p]=3 X_itp[,,p] has a spatially varying linear effect on linear predictor z for category c
slim.output	TRUE or FALSE, if true then vast_output only contains idx and/or idx.se, fit.time, mgc
strata.sp	[Optional] If present, a shapefile containing the strata boundaries to calculate the indicies for
enviro	[Optional] If present, a named-list of length two is required: "formula" is a character string that can be coerced to a formula using stats::as.formula, and "covariate_data" is a data frame with the following columns - Year, Lon, Lat, covariates...
n_xLOC	[Optional] If present, these are the locations used to define the spatial mesh. This is defined as a matrix with two columns 'Lon' and 'Lat' giving the locations for the knots.
newton_steps	An integer value (default is 3), the number of newton steps to take after optimization. Leads to a better mgc but is slow. Setting this to zero turns off this feature.
n.boot	If greater than zero (default is 250), this is the number of replicates to draw from a parametric bootstrap to generate new data from which to calculate DHARMa residuals.

Value

Named list "vast_output"

idx

the index within each strata if strata is provided

idx.se

the associated se for the index

Opt

the diagnostics from the model run

Report

the objects estimated and calculated by VAST

Sdreport

the report generated by ADREPORT

TmbData

the data passed to TMB and used to fit the model

Extrapolation_List

the extrapolation list

fit.time

the time to run the function

MapDetails_List

the map details to make additional plots

PacificCommunity/ofp-sam-vast-utils documentation built on Nov. 2, 2020, 7:05 p.m.