[ADMB Users] Are parameters scaled for optimization.

dave fournier davef at otter-rsch.com
Wed Jan 8 07:51:44 PST 2014

  Bounded initial parameters  are bounded to lie between -1 and 1 from the
  point of view of the function minimizer (sort of).  This has nothing 
to do with what the
optimal scaling for the parameter might be.  To change the scaling one 
can use


member function. set_scalefactor  has been  available for dvariable 
types for a long time,
but was not fully implemented for random effects.  It is now implemented for
init numbers and init vectors  both bounded and unbounded.

Note the unbounded variables are not bounded from the point of view of the
function minimizer   (neither are bounded ones strictly speaking).

Anyway I think there is a confusion here between bounding and optimal
scaling. They are really separate issues.

More information about the Users mailing list