Random submodel

Random terms in the model

The !r random terms in the model formula
  • comprise random covariates, factors and interactions including special functions and reserved words,
  • involve an initial non-zero variance component or ratio (relative to the residual variance) default 0.1; the initial value can be specified after the model term or
  • if the variance structure is not scaled identity, it must be defined using a G structure definition.
  • an initial value of its variance (ratio) may be followed by a !GP (keep positive, the default), !GU (unrestricted) or !GF (fixed) qualifier,
     wwt ~ mu Year age !r sire 0.15 !GP
    
  • use !{ and !} to group model terms that may not be reordered. Normally ASReml will reorder the model terms in the sparse equations - putting smaller terms first to speed up calculations. However, the order must be preserved if the user defines a structure for a term which also covers the following term(s) (a way of defining a covariance structure across model terms). Grouping is specifically required if the model terms are of differing sizes (number of effects). For example, for traits weaningwight and yearlingwight, an animal model with maternal weaning weight should specify model terms
    !{ Trait.animal at(Trait,1).dam !}
    when fitting a genetic covariance between the direct and maternal effects.

    Return to start