Spectral Two-Kernel REML

Estimates variance components for a two-kernel mixed model using Kang et al. (2008) spectral profiling over a convex kernel mixture weight.

Usage

hbr_reml_two_kernel(
  y,
  X,
  G_R,
  K_H = NULL,
  w_grid = seq(0, 1, by = 0.1),
  refine = TRUE
)

Arguments

y

Numeric vector of phenotypes.

X

Matrix of fixed-effect covariates.

G_R

Square relationship matrix (background polygenic kernel).

K_H

Square relationship matrix (feature haplotype kernel) or NULL.

w_grid

Numeric grid for kernel mixing weight w in [0, 1].

refine

Logical. Perform golden section search refinement around best grid point.

Value

List containing `w`, `s2g`, `s2e`, `s2R`, `s2H`, and `negll`.

Examples

set.seed(1)
n <- 30
X <- cbind(1, rnorm(n))
Z <- matrix(rnorm(n * 20), n, 20)
G <- tcrossprod(Z) / 20
y <- rnorm(n)
hbr_reml_two_kernel(y, X, G)
#> $w
#> [1] 0
#> 
#> $s2g
#> [1] 5.243438e-05
#> 
#> $s2e
#> [1] 1.15485
#> 
#> $s2R
#> [1] 5.243438e-05
#> 
#> $s2H
#> [1] 0
#> 
#> $negll
#> [1] 41.74652
#>