Utility function performing Newton-Raphson algorithm updates for penLik.EMNewton
NRupdate(f, starts, gradient, hessian, ..., ridge0 = 1e-06,
tolerance = sqrt(.Machine$double.eps),
iter.max = 1500, halving.max = Inf, relative = FALSE,
return.hessian = FALSE, debugging = FALSE)
Objective function to be minimized
A numeric vector of starting values
The gradient function of f
The Hessian function of f
Additional arguments to be passed to f
A small ridge factor; obsolete. Current version uses nearPD
to stabilize hessian
A small numeric scalar giving the convergence criterion
Maximum number of iterations
Maximum number of step-halfing
A logical scalar indicating if relative convergence should be checked.
A logical scalar indicating if the final Hessian matrix is returned.
A logical scalar indicating if the debuggging mode of the code should be run.
A numeric vector of updated parameters, with attributes
'objective'
The final evaluated objective function
'gradient'
The final gradient vector
'iter'
The number of iterations
'hessian'
The final Hessian matrix, only available if return.hessian=TRUE
.
Long Qu, Dan Nettleton, Jack Dekkers (2012) A hierarchical semiparametric model for incorporating inter-gene relationship information for analysis of genomic data. Biometrics, 68(4):1168-1177