Parametric bootstrap for an asreml model. — bootstrap

Simulate \(\hat{y} \sim N(X\hat{\beta}, V)\) according to the current asreml fit and then refit to obtain the targeted statistics.

Usage

bootstrap_asreml(
  model,
  FUN,
  nsim = 1,
  use.u = FALSE,
  source = list(),
  seed = NULL,
  ...
)

Arguments

model: An asreml fitted model. Must be fitted with model.frame = TRUE.
FUN: A function with signature function(fit) returning a scalar (the statistic to bootstrap).
nsim: Integer. Number of bootstrap replicates.
use.u: A logical indicating whether to resample random effects, or only resample residuals.
source: The known genomic relationship matrix (GRM) used in model fitted using asreml::vm(), provided as a named list. When not provided (an empty list by default), the GRM variable used for vm calling will be searched in the global environment.
seed: Optional integer seed for reproducibility.
...: Additional arguments passed to boot::boot().

Value

A boot object.

Details

Fits parametric bootstrap replicates for an asreml model by:

Extracting the fixed-effect fit yhat = X * beta.
Extracting V = Var(y) on the observation scale,
Simulating new responses y* = yhat + L %*% z where L is a Cholesky factor of V,
Refitting the same asreml call on each simulated dataset,
Returning a boot object.

Examples

if (FALSE) { # \dontrun{
  lettuce_subset <- lettuce_phenotypes |> subset(loc == "L2")
  lettuce_asreml <- asreml(
    fixed = y ~ rep * pseudo_var1,
    random = ~gen,
    sparse = ~pseudo_var2,
    data = lettuce_subset,
    trace = FALSE
  )

  b <- bootstrap_asreml(
    lettuce_asreml,
    R = 200,
    statistic = function(fit) coef(fit)$fixed["(Intercept)", "effect"],
    seed = 1
  )
  boot::boot.ci(b, type = "perc")
 } # }