INLAPPSealsPoissonReg: Seal counting using INLA and Poisson regression formulation

In PointProcess/SealPupProduction-JRSSC-code:

Description

Approximates a Log-Gaussian Cox process with a regression formulation and fits seal counts from transects to it. Samples from the the fitted model to compute the posterior predictive distribution of the total number of seals within an area which the transects are spanning

Usage

INLAPPSealsPoissonReg(
  resultsBaseFolder = "./Results/PoissonReg",
  sealPhotoDataFile = system.file("extdata", "processed", "OriginalSealsKm.rds", package
    = "SealCoxProcess"),
  sealTransectDataFile = system.file("extdata", "processed",
    "OigardTablesTransformed.rds", package = "SealCoxProcess"),
  satelliteDataFolder = system.file("extdata", "processed", package = "SealCoxProcess"),
  sealType = "hooded",
  spatial = TRUE,
  covariate.fitting = "linear",
  covariates.type = "band1",
  additional.iid.term = FALSE,
  noSamp = 5000,
  standardMesh = FALSE,
  convHullVar.convex = -0.15,
  convHullVar.concave = convHullVar.convex,
  convHullVar.resolution = c(120, 120),
  meshVar.max.edge = c(2, 10),
  meshVar.offset = 6,
  meshVar.cutoff = c(0.195, 0.23),
  y.cutoff.boundary = -60,
  Matern.alpha = 2,
  grid.pixelsize = 0.2,
  INLA.theta.startval = NULL,
  INLA.verbose = FALSE,
  parallelize.numCores = 10,
  parallelize.noSplits = parallelize.numCores,
  poisson.maxEvals = 5 * 10^5,
  results.CI.level = 0.95,
  additional_comment = "",
  save.data = TRUE,
  delete.temp = TRUE,
  testing = FALSE,
  leaveOutTransect = FALSE,
  transectAsCountDomain = TRUE,
  CVFold = list(0, 0, 0),
  INLA.constr = TRUE
)

Arguments

resultsBaseFolder	String, indicating the path to the base folder where all results are to be stored
sealPhotoDataFile	String, indicating the file where seal photo data are stored
sealTransectDataFile	String, indicating the file where seal transect data are stored
satelliteDataFolder	String, indicating the path to the folder where the satellite data are stored (in files on the form "cov_grid_[covariates.type].rds")
sealType	String, indicating which seal type to model "hooded" (default as it is quicker), or "harps"
spatial	Logical indicating whether a spatial spde model should be used
covariate.fitting	String, indicating how to model covariates. "no", "linear", "quadratic" (default), "linAndLog" or "nonlinear"
covariates.type	String equal to "band1" or "band2" indicating which of the bands from the satellite data should be used.
additional.iid.term	Logical, indicating whether to include an additional iid (Gaussian) term in the latent field specification. FALSE is default
noSamp	Numeric, indicating the number of samples from posterior model to use when computing the posterior predictive distribution. 5000 (default) typically sufficient.
standardMesh	Logical indicating whether a standard mesh only fixing mesh points at the photo centers should be created. Defaults to FALSE indicating that the special mesh with Voronoi tesselations corresponding to the photos should be used.
convHullVar.convex	Numeric, corrresponding to the convex parameter of inla.nonvonvex.hull. See ?inla.nonconvex.hull for details
convHullVar.concave	Numeric, corrresponding to the concave parameter of inla.nonvonvex.hull. See ?inla.nonconvex.hull for details
convHullVar.resolution	Numeric vector of length 2, corrresponding to the resolution parameter of inla.nonvonvex.hull. See ?inla.nonconvex.hull for details
meshVar.max.edge	Numeric, corrresponding to the max.edge parameter of inla.mesh.2d. Smaller values gives smaller triangles outside triangles. See ?inla.mesh.2d for details
meshVar.offset	Numeric, corrresponding to the offset parameter of inla.mesh.2d. See ?inla.mesh.2d for details
meshVar.cutoff	Numeric vector, where the indicies corrresponds to the cutoff parameter of inla.mesh.2d. See ?inla.mesh.2d for details
y.cutoff.boundary	Numeric vector deciding the y-values which should be the coundary for the different cutoff-values. NULL means the first is applied everywhere
Matern.alpha	Numeric, corresponding to the alpha parameter in the Matern covariance function (2 is the default)
grid.pixelsize	Numeric, denoting the size (in km, in both x- and y-direction) of each pixel of the grid being used
INLA.theta.startval	List containing the start values for the theta parameter in the INLA run. (NULL indicates that automatic start values should be used, and is the default)
INLA.verbose	Logical, indicating whether verbose printing should be used when running the actual inla function
parallelize.numCores	Numeric, corresponding to the number of cores any parallelization should be run at
parallelize.noSplits	Numeric, deciding how many sublists samp should be splitted into. Should be a multiple of parallelize.numCores for the highest efficiency. The larger number the less memory is used (and longer time).
poisson.maxEvals	Numeric, corresponding to maximum number of points the Poisson distribution should be evaluated at (a much smaller number is typically used)
results.CI.level	Numeric, denoting the confidence/credibility degree to use in the final credibility interval for the total number of counts
additional_comment	String, where any comments related to the actual run can be given
save.data	Logical, indicating whether input variables, data and results should be saved (TRUE is default)
delete.temp	Logical, indicating whether temporary stored samples and associated eta grids should be deleted
testing	Logical, indicating whether the testing parts of this function should be used
leaveOutTransect	The transect number for the transect to remove from the modelling fitting and then fit (defaults to FALSE, i.e. none)
transectAsCountDomain	Logical indicating whether the transects should be used as counting domain when they are left out of the modeling (does not apply if leaveOutTransect=F)
CVFold	List of size 3. The is a vector indicating the fold number(s) to remove; the second a single number with the total number of folds; the third is a single number indicates whether it should be a fixed (1) "sample" based on the transect number, or random (0) sample. Defaults to c(0,0,0) which means no cross validation is performed
INLA.constr	Logical deciding whether the SPDE model should be fitted with the constaint that it integrates to zero or not.

Value

Nothing really, but saves results in the resultsBaseFolder

PointProcess/SealPupProduction-JRSSC-code documentation built on Jan. 27, 2020, 10:06 p.m.