msss_fit: Create R1 knots on a grid a specificed number of knots wide

In daktx2/MSSS: Fits multiresolution spatial models with spatially varying resolution

Description

Create R1 knots on a grid a specificed number of knots wide

Usage

msss_fit(locations, yy, knots_r1, spatial_dimension, maxiters = 100,
  cores = 1, design_mat = NULL, stopping_rule = "maxiter",
  g_method = 1, a_pi = 1, b_pi = 5, a_g = 3, kernel_width = 1.5,
  nu = 1, pi_method = 1, R1_prior = NULL, Kernel_type = "bezier")

Arguments

locations	locations of spatial observations.
yy	observation at each location
knots_r1	locations of R1 knots, equally spaced grid, output of r1_create
spatial_dimension	dimension of the locations, should be 1 or 2
maxiters	number of iterations, default is 100
cores	number of cores for parallel processing, ~20 is optimal
design_mat	fixed effects, at least an intercept is recommended
stopping_rule	'maxiter' is default, could be 'log likelihood' if you want it to exit early when finding no new models
g_method	1 for hyper g, 1 i sdefault
a_pi	parameter 1 for beta binomial sparsity if pi_method=1, or numerator if pi_method=2, 1 is default
b_pi	parameter 2 for beta binomial sparsity if pi_method=1, or denominator if pi_method=2, 5 is default
a_g	hyper g prior parameter, 3 is default, 2-4 recommended
kernel_width	bezier kernel width parameter, should be 1.5 or greater to be sensible, 1.5 is default
nu	bezier smoothness parameter, default is 1
pi_method	beta binomial is 1 (default), binomial (fixed pi) is 2
R1_prior	covariance matrix for R1 knots, default is NULL
Kernel_type	either bezier or wendland

Value

large list, exists to be processed by the function msss_predict forsummaries and predictions

daktx2/MSSS documentation built on May 24, 2020, 4:28 a.m.