Numerical Issues Involved in Inverting Hessian Matrices
Jeff Gill, Gary King · Wiley series in probability and statistics · 2003
Chapter four provides some methods for inverting singular Hessians such that we are sometimes able to extract information in a convenient format from problematic likelihood functions or posterior distributions without respecification. Topics include: developing a solution using Bayesian simulation tools problem in detail: noninvertible Hessians A generalized inverse/generalized Cholesky solution generalized inverse generalized Cholesky Gill/Murray Cholesky factorization Schnabel/Eskow Cholesky factorization importance sampling and sampling importance resampling drawing from the singular normal aliasing ridge regression derivative approach bootstrapping