fitrvine: Fit a Vine Model
In vincenzocoia/copsupp: Vine Copulas made easy

Description Usage Arguments Details Value Examples

Fits a joint distribution for the data using a vine copula. The vine array is first chosen using the minimum spanning tree algorithm using the function gausstrvine.mst in the CopulaModel package, then the pairwise copula models are chosen and fit using RVineCopSelect in the VineCopula package.

1	fitrvine(dat, layer = layeropts(1:ncol(dat)), basevine = NULL, ...)

`dat`	Matrix of data having Uniform marginal distributions; columns represent variables, and rows observations.
`...`	Other arguments to pass to `VineCopula::RVineCopSelect`.
`rv`	Object of class "rvine", representing a pre-specification of the vine to fit. Or, `NULL` for fully unspecified. See details.
`var`	Vector of integers specifying the column numbers of `dat` to fit a model to. Default is all variables.
`ntrunc`	Integer, either `1, 2, ...,ncol(dat)-1`, of the truncation level of the vine to be fit.
`families`	A vector of copula family names to try fitting (will also consider their rotations/reflections). Limited to those families available in `VineCopula` package, listed in `BiCopSelect`.

If you want to specify parts of the vine, then specify them in an "rvine" object (see rvine):

Your first option is to specify the vine array A.
If the vine array is specified, then you can specify some or all of the copula families by putting them in copmat. Leave unspecified edges as NA.
If there are copula families specified, you can specify parameters for those families by putting the parameters in cparmat. Unspecified parameters should be NA.

For parts of the vine that are unspecified, you have some fitting options:

If you didn't specify a vine array A, you can select which variables (column numbers of dat) you'd like to fit through the var argument. You can also select the truncation level of the vine array through the argument ntrunc.
If you left some copula families unspecified, you can indicate the candidate families in the families argument.

A "fitrvine" object, which has class c("fitrvine", "rvine"), which is a named list of the following:

$A: Vine array, truncated to ntrunc.
$copmat: ntrunc x ncol(A) upper-triangular matrix of copula model names.
$cparmat: ntrunc x ncol(A) upper-triangular matrix of copula parameters. Each entry is a list of length one containing the vector of copula parameters for that copula family.
$dat: The inputted data matrix, dat.
$aic: The AIC of the fitted model.
$bic: The BIC of the fitted model.
$nllh: The negative log likelihood of the fitted model.
$covmat: Covariance matrix of the fitted parameters (in reading-order of the parameters, i.e. c(t(cparmat)[lower.tri(t(cparmat))], recursive=TRUE)).

## Get some simulated data:
set.seed(152)
ntrunc <- 2
d <- 4
A0 <- truncvarray(CopulaModel::Dvinearray(d), ntrunc)
copmat0 <- makeuppertri("frk", ntrunc, d, "")
cparmat0 <- makeuppertri(3, ntrunc, d)
dat <- fvinesim(100, A0, copmat0, cparmat0)

## Fit a model to the data:
fit.rvine(dat, ntrunc=ntrunc)
fit.rvine(dat, c(4, 2, 3))