Description
asv.asreml computes the approximate stratum variances, degrees of freedom and coefficients for simple variance component models.
Usage
## S3 method for class 'asreml'
asv(object, order = c("as.is", "noeff"), vvp = FALSE, ...)
Arguments
object |
An object of class asreml. |
order |
The sequence in which to consider the random terms: may be one of “as.is” or “noeff” for the order as given in the asreml object, or ordered by increasing number of effects, respectively. The default is “as.is“. Note that the default ordering for random terms in asreml is “noeff“. |
vvp |
If TRUE (the default is FALSE) the inverse of the average information matrix for the variance components (sigma scale) is returned. |
| … | Any additional arguments. |
Details
Approximate stratum variances and degrees of freedom for simple variance components models are returned. For the linear mixed-effects model, it is often possible to consider a natural ordering of the variance component parameters, including the residual. Based on an idea due to Thompson, 1980, asreml computes approximate stratum degrees of freedom and stratum variances by a modified Cholesky diagonalisation of the average information matrix.
Value
If vvp is FALSE, a rectangular matrix containing the degrees of freedom, approximate variances and the component coefficients for each random term. If vvp is TRUE, a list of length two containing the matrix of approximate stratum variances and the inverse average information matrix of the variance components in list components asv and vvp, respectively.
References
Thompson R (1980). “Maximum likelihood estimation of variance components.” Series Statistics, 11(4), pp. 545-561.