Calculate profile traces for fitted models.
# S3 method for evd
profile(fitted, which = names(fitted$estimate), conf = 0.999,
mesh = fitted$std.err[which]/4, xmin = rep(-Inf, length(which)),
xmax = rep(Inf, length(which)), convergence = FALSE, method = "BFGS",
control = list(maxit = 500), ...)
An object of class "profile.evd"
, which is a list with an
element for each parameter being profiled. The elements are
matrices. The first column contains the values of the profiled
parameter. The second column contains profile deviances. The
remaining columns contain the constrained maximum likelihood
estimates for the remaining model parameters. For calculation of
profile confidence intervals, use the confint.profile.evd
function.
An object of class "evd"
.
A character vector giving the model parameters that are to be profiled. By default, all parameters are profiled.
Controls the range over which the parameters are profiled.
The profile trace is constructed so that (assuming the usual
asymptotic properties hold) profile confidence intervals with
confidence coefficients conf
or less can be derived from it.
A numeric vector containing one value for each
parameter in which
. The values represent the
distance between the points profiled. By default mesh
is
one quarter of the standard errors. If the fitted object does not
contain standard errors the argument must be specified.
The argument should also be specified when an estimator is
on or close to a parameter boundary, since the approximated
``standard error'' will then be close to zero.
Numeric vectors containing one value for each
parameter in which
. Each value represents the theoretical
lower/upper bound of the corresponding parameter.
The arguments are needed only when a parameter has a
lower/upper bound at which the likelihood is non-zero. Do not
use these arguments to specify plotting ranges in a subsequent
plot (as they are used in the calculation of profile confidence
intervals); to do this use xlim
in the call to plot
.
Logical; print convergence code after each
optimization? (A warning is given for each non-zero convergence
code, irrespective of the value of convergence
.)
The optimization method.
Passed to optim
. See optim
for
details.
Ignored.
confint.profile.evd
, profile2d.evd
,
plot.profile.evd
uvdata <- rgev(100, loc = 0.13, scale = 1.1, shape = 0.2)
M1 <- fgev(uvdata)
if (FALSE) M1P <- profile(M1)
if (FALSE) par(mfrow = c(2,2))
if (FALSE) cint <- plot(M1P)
if (FALSE) cint
Run the code above in your browser using DataLab