Fit a low-rank LMM on raw genotypes (no explicit scaling).

biglmmg(
  y,
  X,
  G,
  cols = seq(ncol(G)),
  M = length(cols),
  K = NULL,
  REML = TRUE,
  compute_mult = TRUE,
  verbose = 0
)

Arguments

y	A vector of trait values (quantitative trait).
X	A matrix of covariates. The default value is matrix of ones for the intercept (one column).
G	A FBM matrix of genotypes. Missing values are not handled.
cols	A vector of columns of G to be used in the model. By default, all columns of G are used.
M	A scalar for normalization of the genetic relationship matrix: GRM = Z'Z / M, where Z is a matrix of standardized genotypes. By default, M = length(cols).
K	A matrix with the pre-computed cross-product Z'Z / M. By default, K = NULL, that means K is pre-computed inside the function.
REML	A boolean specifying the likelihood function, REML or ML.
compute_mult	A boolean enabling the computation of the effective sample size, when only model fitting is needed. The default value is TRUE.
verbose	The verbose level. The default value is 0 (verbose).

Details

The linear mixed model (LMM) is: y_i = X_i b + u_i + e_i, where

u ~ N(0, s2*h2*G) and e ~ N(0, s2 I)

var(y) = V = s2 * (h2*G + I)

Examples

G <- attach_example200() # load simulated genotypes
G
#> A Filebacked Big Matrix of type 'code 256' with 1500 rows and 200 columns.
G[1:5, 1:10]
#>      [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10]
#> [1,]    2    1    0    2    1    1    1    1    2     1
#> [2,]    1    2    1    1    2    1    1    2    1     1
#> [3,]    2    2    1    1    1    1    1    0    2     0
#> [4,]    0    2    0    2    1    0    0    1    0     2
#> [5,]    1    2    0    2    2    1    0    0    1     0

y <- rnorm(nrow(G)) # generate a random phenotype
head(y)
#> [1]  0.51242695 -1.86301149 -0.52201251 -0.05260191  0.54299634 -0.91407483

mod <- biglmmg(y, G = G)
mod$gamma # estimated h2
#> [1] 6.610696e-05