optimizer_biv_ext_gauss_model: Optimize the coefficients of the best fitted linear models...

In SpatialModelsZAMG: Spatial Modeling of Extreme Events

Description

this function optimizes the coefficients of the best fitted linear models
(from the function model_selection) via bivariate Extremal-Gaussian model
(= Extremal-t model with fixed nu = 1) and with composite likelihood inference

Usage

optimizer_biv_ext_gauss_model(sd_m_select, swe_m_select,
                              method = "ucminf",
                              follow.on = FALSE,
                              itnmax = NULL,
                              printParam = FALSE)

Arguments

sd_m_select	this input should be a list including max_data, covariables and models as in the output of the function model_selection
swe_m_select	this input should be a list including max_data, covariables and models as in the output of the function model_selection
method	optimization method(s) for external function optimx, this can also be a vector, but optimization might take several hours for one method! possible methods are: Nelder-Mead, BFGS, CG, L-BFGS-B, nlm, nlminb, spg, ucminf (default), newuoa, bobyqa, nmkb, hjkb, Rcgmin, Rvmmin
follow.on	logical value; if TRUE, and there are multiple methods, then the last set of coefficients from one method is used as the starting set for the next default is FALSE
itnmax	if provided as a vector of the same length as the length of method, this gives the maximum number of iterations or function values for the corresponding method. if a single number is provided, this will be used for all methods
printParam	logical value; if TRUE, the GEV parameters during the optimization are printed. this might be useful to check the proper functioning of the optimization (shape parameter should be approximately between -0.5 and 0.5) default is FALSE

Value

a list with

complete_sd_max_data	a completed submatrix of the max_data matrix for sd, where the completation was performed via function complete of the package mice. each row corresponds to one station (unchanged), columns (years) were chosen such that less than 50% of the original column-entries were NA's
complete_swe_max_data	completed submatrix of the max_data matrix for swe, where the completation was performed via function complete of the package mice. each row corresponds to one station (unchanged), columns (years) were chosen such that less than 50% of the original column-entries were NA's
summary	a summary of the optimization results, including an information message whether the optimization was successful or not and which method delivered the best coefficients
coefficients	a list with the optimized coefficients. containing: sd_coeff (sd_loccoeff, sd_scalecoeff, sd_shapecoeff), swe_coeff (swe_loccoeff, swe_scalecoeff, swe_shapecoeff), cor_coeff (alpha, sd_kappa, swe_kappa, rho) and all_coeff

References

Genton, M.G. & Padoan, S.A. & Sang, H. (2015): Multivariate max-stable spatial processes. Biometrika 102(1): 215-230.

http://repository.kaust.edu.sa/kaust/bitstream/10754/552385/1/2013.GPS.Biometrika.Rev_14.pdf

See Also

model_selection, optimizer_biv_ext_t_model, optimizer_biv_hr_model

SpatialModelsZAMG documentation built on Nov. 11, 2019, 3 p.m.