likelihood.gradient.augmented | R Documentation |
This function uses the
likelihood.gradient
function and adds the linear
constraints used in fit.gev
and
fit.gpd
to produce the augmented Lagrangian
version of the GEV or GP negative log-likelihood function.
likelihood.gradient.augmented(parameters, x.in, model = c("gev", "gpd"), lagrangian.multiplier = rep(0, length(x.in) + 2), penalty.parameter = 1000)
parameters |
Numerical vector containing the location, scale,
and shape parameters for the GEV or the scale and shape
parameters for the GP. If NULL,
|
x.in |
Time series of class xts. |
model |
String determining whether to calculate the initial parameters of the GEV ("gev") or GP ("gpd") function. Default = "gev". |
lagrangian.multiplier |
Lagrangian multipliers used to weight the linear contribution of the constraints. In most cases all of them are zero, since optimization of the GEV/GP likelihood usually doesn't take place inside a region of constraint violations. When supplying this parameter it has to have the same length as present number of constraints: number of points in x.in + 2, with the last two constraints handling the lower bound of the scale and the shape parameter. Default = 0 for all constraints. |
penalty.parameter |
Penalty parameter used to weight the quadratic contribution of the constraints. In the end of a typical constrained GEV or GP optimization this parameter is 1000. Default = 1000. |
A convenience function not used by the fitting routines.
It is only meant to work with constant parameters and no covariates.
Numerical value of the gradient of the augmented negative log likelihood.
Philipp Mueller
Other optimization: fit.gev.default
,
fit.gev.list
, fit.gev.xts
,
fit.gev
, fit.gpd.default
,
fit.gpd.list
, fit.gpd.xts
,
fit.gpd
,
likelihood.augmented
,
likelihood.gradient
,
likelihood.initials
,
likelihood
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.