MATLAB Answers

how to assess the determinability of a model optimised using lsqnonlin?

2 views (last 30 days)
Aisling on 20 Jan 2012
I have used lsqnonlin to find the best-fit parameters of a model. Now I would like to examine how reliable the optimisation was: I would like to assess the 'determinability' of the model using the Hessian curvature matrix.
In short I need 1) the determinant of the Hessian (d-optimality criterion) 2) the condition number of the hessian (the ratio between its smallest and largest eigenvalues) 3) the m-optimality criterion
The problem is lsqnonlin does not return the Hessian (it does return the Jacobian - can I get the Hessian from that?) I could use hessian.m in the derivest toolbox on the file exchange but it seems to only return one point at a time e.g solve when (x,y) = (1,1). If I looped through for every (x,y) point that I have I don't know how to construct the matrix after that.....
I'm no expert on this stuff so please don't crucify me but point me in the right direction instead!


Sign in to comment.

Answers (0)