Compute LD Matrix: Standard r^2 or Kinship-Adjusted rV^2

Unified dispatcher. With method = "r2" (default and recommended for large datasets) it calls the C++ Armadillo back-end for fast standard squared Pearson correlations. With method = "rV2" it applies kinship whitening first and then the same C++ correlation kernel.

Usage

compute_ld(X, method = c("r2", "rV2"), digits = -1L, n_threads = 1L)

Arguments

X

NumericMatrix (individuals x SNPs). For method = "r2": raw or mean-centred genotypes (0/1/2). For method = "rV2": the pre-whitened matrix \(V^{-1/2} \tilde{G}\) produced by prepare_geno().

method

Character. "r2" (default) or "rV2".

digits

Integer. Rounding precision. -1 skips rounding.

n_threads

Integer. OpenMP threads for the C++ kernel. Default 1. Use parallel::detectCores() to choose automatically.

Value

Symmetric p x p numeric matrix, diagonal 0, values in [0, 1].

See also

compute_r2, compute_rV2, prepare_geno, Big_LD()

Examples

if (FALSE) { # \dontrun{
# Internal function -- use compute_r2() or compute_rV2() instead
set.seed(1)
G <- matrix(sample(0:2, 60 * 20, replace = TRUE), 60, 20)
ld_r2  <- LDxBlocks:::compute_ld(G, method = "r2")
range(ld_r2)
} # }