Sigma parameterization for simple models
!SIGMAPAR
!SIGMAPAR
is used to override the default behaviour of ASReml which is
to fit univariate models using the GAMMA parameterization.
Under the gamma paraterization, variance parameters or expressed
relative to the residual variance. So for example, if the residual variance was
100 and a random term variance component was 10, the variance parameter would be estimated
as 0.1.
!SIGMAPAR
causes the variance parameters to be estimated directly on the variance scale.
!SIGMAPAR may be specified among the data file line qualifiers,
on the model line before the ~ symbol, or on the
residual line.
It is sometimes needed in conjunction with the
!ASUV qualifier
where you have multivariate data but are fitting a residual structure
other than US, for example DIAG. If the variances are included in this other structure,
you will need to specify !SIGMAPAR.
Return to index