user_vs {asreml}R Documentation

User-defined variance models or structures

Description

Specify an external function that provides a variance-covariance matrix, or its inverse.

Usage

own(obj, fun = "myowng", is.variance = TRUE)

Arguments

obj

A factor in data.

fun

The name (as a character string) of an R function to compute the variance matrix and its derivative(s). This function must accept a single argument:

  • order: a scalar giving the dimension of the structure being defined,

and return a list containing the matrix (the variance matrix or its inverse) and the corresponding derivative matrices. The variance matrix may be a dense matrix class object, a vector being the lower triangle in row-major order, or a three-column matrix in coordinate form in row-major order.

This object may have an attribute INVERSE, a logical scalar identifying the structure as a variance matrices or its inverse. If INVERSE attribute is absent the default is FALSE.

is.variance

If TRUE, then own assumes the resulting structure is a variance matrix, otherwise a correlation matrix is assumed. This only affects the default scaling factor (default = TRUE).

Details

The own variance model allows users to specify external variance structures. This requires the user to provide an R function that accepts a single argument, the leading dimension of the structure, and forms the variance matrix (or its inverse). The R function may invoke compiled code if necessary.

Examples

## Not run: 
## An own ar1 function
ar1.asr <- asreml(yield ~ Variety,
                  residual = ~ar1(Row):own(Column, "ar1", 0.1, "R"),
                  data = shf)

## where the function "ar1" is defined as:

## ar1 <- function(order, kappa) {
##   t <- 1:order
##   H <- abs(outer(t, t, "-"))
##   V <- kappa^H
##   ## derivative
##   dV <- H*(kappa^(H-1))
##   return(list(V, dV))
## }

## End(Not run)
[Package asreml version 4.2.0.480 Index]