SOIR: Scalar-on-Image regression in STAR framework using sarim

In RaphaelRe/SOIR: Scalar-on-Image regression within Sarim

Description

Returns a sx-object for Scalar-on-Image regression within sarim function

Usage

SOIR(images, dimension = rep(sqrt(ncol(images)), 2),
  neighbours = c("2dfirst", "2dsecond", "2dallfirst", "3dfirst",
  "3dsecond", "3dallfirst"), solver = c("rue", "lanczos"),
  demean = FALSE, add_diag = NULL, ...)

Arguments

images	The vectorized images (columnwise)
dimension	Vector which defines the dimension of the images
neighbours	String defining the type of neighbourhood of underlying GMRF
solver	String defining the solve which should be used
demean	Logical, definign wether the mena of images should be substraced
add_diag	numeric value, default is NULL. For numeric stability one can add a value to main diagonal to structure matrix
...	Further arguments for the underlying function Sarim::sx(), most relevant are ka_a, ka_b and ka_start

Value

sx-object with images, structure matrix and so on.

Examples

### simple call:
ims <- matrix(1:(100*32*32), nrow = 100, ncol = 32*32)
SOIR(ims)

### visualization of the structure matrix
plot(SOIR(ims))

### Call within Sarim:
beta <- rnorm(32*32)
y <- ims  %*% beta + rnorm(100, sd = 5)
mod <- Sarim::sarim(y ~ SOIR(ims, add_diag = 0.1, ), nIter = 10)

RaphaelRe/SOIR documentation built on May 28, 2019, 12:19 p.m.