nlminb
nlminb
.
nlminb.control(eval.max, iter.max, abs.tol, rel.tol, x.tol, step.min, step.max, scale.init, scale.upd, scale.fac, scale.tol, scale.mod, sing.tol, sing.step, diff.g)
scale
. If
scale.init
>= 0, all components of the scale vector will be initialized to
scale.init
before updating. If
scale.init
< 0, the scale vector will not
be initialized internally. Default : 0.0
scale
at the beginning of
an iteration. Each component of
scale
is set either to the norm of the
corresponding diagonal element of the Hessian matrix or else to
scale.fac
times its current value, whichever is larger. Default : 0.6
scale
during
automatic updating. If a component of
scale
becomes smaller than
scale.tol
in magnitude, it is set to
scale.mod
.
Default :
scale.tol
= 10^(-6).
scale
during
automatic updating. If a component of
scale
becomes smaller than
scale.tol
in magnitude, it is set to
scale.mod
.
Default :
scale.mod
= 1.0.
sing.tol
= max( 10^(-10), .Machine$double.eps^(2/3))
sing.step
= 1
nlminb
.
Automatic updating for
scale
is available only if
hessian
is supplied.
nlminb.control
would normally be called only to set the
control
parameter
in a call to
nlminb
. Names may not be abbreviated.
For details, see the PORT Mathematical Subroutine Library Manual.
A. T. & T. Bell Laboratories (1984). PORT Mathematical Subroutine Library Manual.
norm.neg.2.11 <- function(theta, y) { length(y) * log(theta[2]) + (1/theta[2]) * sum(y-theta[1])^2 } set.seed(12) my.obs <- rnorm(100, mean=10, sd=2) nlminb(start = c(0,1), obj = norm.neg.2.11, control = nlminb.control(iter.max = 100, x.tol = .0001), lower = c(-Inf, 0), y = my.obs)