Description Usage Arguments Value Warning Author(s) Examples
This function finds the maximum composite likelihood estimate
for multivariate extreme value models based on pairs of data.
It returns a list with the result from optim
along with some of the information passed to the function.
1 2 3 |
dat |
|
u |
|
lambda |
vector of percentage of threshold exceedances (can be omitted) |
N |
total number of observations if some observations falling under all threshold were removed from |
model |
family of multivariate extreme value distribution. |
cscale |
logical indicating common scale for all variables |
cshape |
logical indicating common shape for all variables |
sym |
logical indicating whether equal dependence parameters in the Dirichlet model |
start |
vector of starting value for the optimization |
std.err |
logical indicating whether to calculate standard errors numerically |
method |
optimization method; see optim |
warn.inf |
whether to check the validity of starting values before optimization |
... |
additional arguments (fixed parameters) passed to the |
a list containing the output of optim
, notably
parameter values par
and standard errors se
users should strip the data matrix from full NA
cases or else provide N
.
While these will be ignored in the routine, they are taken into account in the calculation of the total number of
non exceedances.
Leo Belzile, some Cpp code by Adrien de Casto and Leonid Rousniak for ‘hr’ and ‘xstud’ models
1 2 3 4 5 6 7 8 9 10 11 12 13 | set.seed(4)
x <- mev::rmev(n=10000, d=2, param=c(1,2,0.5), model = "negdir")
qu <- apply(x, 2, quantile,probs <- 0.99)
y <- x[which(rowSums(isAbove(x,threshold = qu))>=1),]
fit <- fmvcpot(dat=y, u=qu, model="negdir",lambda=colSums(isAbove(y,qu))/(nrow(x)+1), N=nrow(x),
cscale=TRUE, cshape=TRUE, shape=1, scale=100)
fit$par
x <- mev::rmev(n=10000, d=2, param=c(1,2,0.5), model = "dir")
qu <- apply(x, 2, quantile,probs <- 0.99)
y <- x[which(rowSums(isAbove(x,threshold = qu))>=1),]
fit <- fmvcpot(dat=y, u=qu, model="dir",lambda=colSums(isAbove(y,qu))/(nrow(x)+1), N=nrow(x),
cscale=TRUE, cshape=TRUE, shape=1, scale=100)
fit$par
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.