glm with Newton Raphson
For canonical links they are the same thing (and the likelihood is log-concave), but in general NR is a poor optimizer without at least step-length adjustment. MASS4 p.445 has an example of using a general optimizer for logistic regression.
On Fri, 19 Nov 2004, Valeska Andreozzi wrote:
Does anyone know if there is a function to find the maximum likelihood estimates of glm using Newton Raphson metodology instead of using IWLS.
Brian D. Ripley, ripley at stats.ox.ac.uk Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UK Fax: +44 1865 272595