rrBlupMethod6 (version 1.2)

rrBlupM6: rrBlupM6 -- RR-BLUP for estimation of marker effects with fixed residual variance

Description

This function implements "Method 6" described in Piepho et al. (2011), a re-parametrization of Kang et al.'s (2008) mixed model formulation to allow for a fixed residual variance when using RR-BLUP for genomwide estimation of marker effects.

Usage

rrBlupM6(y, X = matrix(1,nrow=n,ncol=1), Z, sig2e, chunks = as.integer(1))

Arguments

y

Numeric vector with phenotypic observations (for example the entry means).

X

Design matrix of fixed effects, including the intercept. By default, this is an all 1 column vector for the intercept.

Z

Matrix assigning marker genotypes to phenotypes in y. The dimension of the matrix must be no. phenotypes (rows) times no. markers (columns). The coding must be 1 and -1 for the two homozygous genotypes.

sig2e

The value of the residual variance, numeric vector of length 1.

chunks

Integer giving the number of chunks into which to split the computation of ZZ'. Computing this matrix in chunks might have computational advantages when the number of markers and observations is very large and the available memory is low. The

Value

• A list with three components [object Object],[object Object],[object Object]

Details

Please see Piepho et al. (2011) for details on the re-parametrization and the computation of ZZ' in chunks. Currently only bi-allelic markers are supported.

See Also

rrBlupMethod6

Examples

## simulate a small data set (250 observations, 300 markers)
set.seed(3421475)
N <- 250
M <- 300

Z <- matrix(sample(c(1,-1),N * M, replace = TRUE),
            nrow = N,
            ncol = M)

## marker effects
u <- rnorm(M, 0, sqrt(1/M))

sig2e <- 1
y <- Z %*% u + rnorm(N,0,sqrt(sig2e))

out <- rrBlupM6(Z = Z, y = y, sig2e = sig2e)

## estimated marker effect variance
out$sig2u



## accuracy in validation set
Z.val <- matrix(sample(c(1,-1),N * M, replace = TRUE),
                nrow = N,
                ncol = M)

cor(Z.val%*% out$uhat, Z.val%*% u)

## compute ZZ' in chunks
out.chunks <- rrBlupM6(Z = Z, y = y, sig2e = sig2e, chunks = 6)

cor(Z.val%*% out.chunks$uhat, Z.val%*% u)