Optional input
==============

.. _optional:

Optional input TASK=1 (AIREML)
------------------------------

.. code-block:: Dmuai

  DMUAI can read an optional file containing parameters to control the maximization process. 
  The format is five lines with one parameter per line in the sequence shown in the table below. 
  The values shown are the default values.

     Value        Name               Description
     ______     ___________  __ ______________________________________________________
     10         EMSTEP        Number of steps before full weight on EM in IMET = 1
     1.0d-7     CONV_NDELTA   Convergence criteria for norm of the update vector
     1.0d-6     CONV_GNORM    Convergence criteria for norm of the gradient vector (AI)


Optional input TASK=2 (MCMC)
----------------------------

.. code-block:: Rjmc

    Mandatory lines
    3 0             !      Igibkode, Itest 
    1111  2111      ! Starting seed for random number generator
    2000  3000 1    ! Burnin, No of rounds, and interleaving
    1  0            ! CV matrix number and prior degree of belief (DF), one line for each CV matrix
                    ! (DF < CV matrix dimension + 2 -> non informative prior)

    Optional model dependent lines for Reaction norm models with unknown environmental co-variate
    1               ! No. of traits times no. of class effects used  as co-variate
                    ! 2 lines per each combinations of no. traits times no. of co-variate
    1 2 1            ! trait (seq. number), E-effect (pos. in model), number of cov
    1 0              ! RN-cov, associated cov

    Additional optional line for Reaction norm models
    0             ! 0: original class eff, 1= standardized std


Optional input TASK=11 (BLUP incore)
------------------------------------

.. code-block:: Dmu4

  DMU4 can compute standard error of selected solutions. This is particularly useful when using one of the
  iterative solvers, because such procedures do not yield standard errors of the solutions because this
  requires the inverse of the coefficient matrix for the mixed model equations. It is, however, possible to
  obtain selected elements of this inverse using the same iterative process.

  It is therefore possible to specify up to 100 equations where the standard error of the solution is desired.
  DMU4 will print a table with the selected solutions and their standard errors, and another table with
  correlations among all solutions. Based on this, various hypotheses can be tested.

  The selected solutions can be specified in the driver file on lines following the $DMU4 keyword line, or in
  a file pointed to by the fn option in the $DMU4 keyword line. The format  is one equation number per line,
  and a max. of 100 lines is possible.

Optional input TASK=12 (BLUP PCG iteration)
-------------------------------------------

.. code-block:: Dmu5

  DMU5 can read optional input consisting of up two lines.

  The first line controls the iteration process by specification of maximum number of iterations and
  convergence criteria. 

  On the next line, the amount of physical RAM (in MB) available can be specified. This is used to optimize
  memory utilization and avoid swapping, by storing as much data in memory as possible.

  If no optional input is given, the following default values are used:

       Maximum round:		1000
       Convergence criteria:	1e-9
       RAM (MB):			2000