Model

The model description is in a number of sections.

TRAITS

For TASK 1, 11 and 12 (DMUAI, DMU4 and DMU5) the line must contain one integer value for number of traits in
the analyses.

For TASK 2 (RJMC) the line must have 5 integer values:
  1: number of traits in the analysis
  2: number of Gaussian traits
  3: number of right censored Gaussian traits
  4: number of ordered categorical traits
  5: number of binary traits

ABSORB

Absorb is not used in release 5.3 or later.

In order to make the shift to release 5.3 easy, absorb line(s) can be still be specified, but will be ignored
during reading the directive file. Therefore, directive files with absorb line(s) with zero(s) (0) for each
trait can still be used.

CLASS Effects

One line for each trait containing several integers:

    1: Specify the real input number for the variable to be analyzed.

    2: Indicator for a weight variable (WV).
       If WV = 0 no weight variable is used.
       If 0 < WV <= # of real input columns), real column WV in the data file is used as weight.
       (For more information on a weighted analysis see the section on weights below).

    3: Specify the number of class effects in the model for this trait (NEF).

4-NEF: Specify the integer input numbers for each class variable in the model for this trait.
       Fixed effects must be given first in each sub model followed by random effects.
       Integer input number 0 can be used to include an intercept in the model.

RANDOM

One line per trait in the same sequence as in the MODEL line. The first number is the number of random
effects in the model for this trait, followed by a numbering of the random factors.
This numbering is used to find random effects that might be correlated such as maternal and direct genetic
effects, which must be specified as two effects on the model line. It is also used to group effects that
are correlated across traits.

REGRES

One line per trait in the same order as in the model statement.

The line starts with the number of regressions in the model for the trait. If no covariables are desired for
this trait, a zero must be specified. On the rest of the line the real input numbers for the covariables must
be specified. Overall regressions can be specified as a single integer number (the real input number for the
covariable). Nested regressions are specified as covariable number followed by the effect number for nesting
in brackets. The figure to specify in bracket is the effect number for the effect where the regression is
nested within, as specified in the model statement for the trait. If the model includes more than one
regression on the same covariable (i.e. overall and nested regressions or regressions nested within several
effects), it can be put into one set of brackets separated by blanks.

The following three specifications are treated identically by the program:

   7 1 2 2(3) 2(4) 3 4 4(3)

   7 1 2 2(3 4) 3 4 4(3)

   7 1 2(0 3 4) 3 4(0 3)

Nesting within effect 0 is the same as an overall regression. All occurrence of a covariable has to be
completed before starting on a new covariable.

NOCOV

A line specifying the number of covariances among residuals that is nonexistent. Such a situation can occur
if the two traits in question are measured on different animals or experimental units. On the following
lines, the traits with non-existing covariances are specified with one set per line.

If no such sets exist in the model, a zero must be specified.