Weighted
========

.. _Weigted:

Weighted
---------

.. code-block:: Weighted

   Wighted analysis can be usefull if the recording for the trait is known to have different residual
   variance, as for instance when analising means of single recordings where the number of recordings
   behind  the mean differ. 

   If no weights are specified (0 as weight variable on the CLASS EFFECT line) a weight of 1.0 is used.
   The weight modifies the residual variance to V = (1/w)Ve, where w is the weight and Ve is the general
   residual variance specified as prior or is the current estimate in a REML analysis. In multivariate
   analysis residual correlations is kept constant when applying weights. 

   Generally, the smaller the weights the less influence of the observation where the weight is applied.