14 January 2015 0 2K Report

The issue is: may we find any apriori (necessary and/or sufficient) conditions on a given nonlinear model: y = f(X,β) + e 

in order to avoid multicollinearity of the approximated Hessian matrix of the error function (i.e. to get the Newton direction accessible during the iterative procedure of optimization) ?

More Pietro Piu's questions See All
Similar questions and discussions