gausstrvine: Best Gaussian truncated vines
In YafeiXu/CopulaModel: CopulaModel: Dependence Modeling with Copulas

Description Usage Arguments Details Value See Also Examples

Best Gaussian truncated d-dimensional vines up to d-2 trees

1 2	gausstrvine(rmat,iprint=F) gausstrvine.nonuniq(rmat,jtrunc=3,eps=1.e-7,iprint=F) # check non-uniqueness

`rmat`	dxd correlation matrix, 4<=d<=8
`iprint`	print flag for intermediate steps (in f90 code)
`jtrunc`	truncation level to check on degree of non-uniqueness
`eps`	tolerance to check on degree of non-uniqueness, default 1e-7

Note that even if the optimal ell-truncated vine is unique, the vine array leading to it is not unique. The output of gausstrvine.nonuniq() is not saved into R variables, so it should be looked at for other truncated vines that lead to the same determinant as the optimal.

`bnum`	d-2 dimensional vector with indices of best vine arrays; can get vine arrays with something like vnum2array(d, bnum[ell])
`logdetmx`	d-1 dimensional vector with max log determinants for truncated vines of order 1,...,d-2; the last entry in position d-1 is the log determinant of rmat
`permmat`	dx(d-2) matrix, with permutation leading to a best ell-truncated vine in column ell
`pcarr`	dxdx(d-2) partial correlation array with matrix of partial correlations in pcarr[,,ell] for the best ell-truncated vine

gausstrvineMST

## Not run: 
rmat=matrix(c(
1.00000,0.69965,0.70477,0.66536,0.65967,
0.69965,1.00000,0.65499,0.61713,0.61202,
0.70477,0.65499,1.00000,0.62967,0.62798,
0.66536,0.61713,0.62967,1.00000,0.57398,
0.65967,0.61202,0.62798,0.57398,1.00000), 5,5)
out=gausstrvine(rmat,iprint=FALSE)
print(out)
outnonuniq=gausstrvine.nonuniq(rmat,jtrunc=3,eps=1.e-7,iprint=TRUE)
# some checks
d=nrow(rmat)
for(ell in 1:(d-2))
{ A=vnum2array(d,out$bnum[ell])
  cat("truncation level ", ell,"\n")
  print(A)
  cat("check on log determinant\n")
  pcmat=out$pcarr[,,ell]
  logdet=sum(log(1-pcmat[1:ell,]^2))
  print(logdet)
}
print(determinant(rmat,log=TRUE)$modulus) 
## End(Not run)