Maximize a log-likelihood by Newton–Raphson with step halving. grad defaults to numerical differences; the Hessian is always numerical (from the gradient). Falls back to Nelder–Mead polishing when Newton stalls.
grad
Optional
Maximize a log-likelihood by Newton–Raphson with step halving.
graddefaults to numerical differences; the Hessian is always numerical (from the gradient). Falls back to Nelder–Mead polishing when Newton stalls.