[ADMB Users] constrained correlation matrices

[dave fournier]

I realized that it is possible to formulate constraint on a correlation 
matrix
of the form

                rho_i,j =0

for some i,j as an unconstrained problem.  To avoid degeneracy a 
quadratic penalty
is added for each constraint.  see the attached example code.

It also appears to provide a solution to this problem

https://stat.ethz.ch/pipermail/r-help/2010-March/230667.html

if you start with the inverse of the correlation matrix instead.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: cc.zip
Type: application/zip
Size: 1137 bytes
Desc: not available
URL: <http://lists.admb-project.org/pipermail/users/attachments/20121121/a93e6313/attachment.zip>