Given a sample of positive definite matrices,
perform numerical maximization of the h-block least square (lscv) or leave-one-out likelihood (lcv) cross-validation criteria using a root search.
bandwidth_optim(
x,
criterion = c("lscv", "lcv"),
kernel = c("Wishart", "smlnorm", "smnorm"),
tol = 1e-04,
h = 1L
)double, the optimal bandwidth up to tol
sample of symmetric matrix observations from which to build the kernel density kernel
optimization criterion, one of lscv for least square cross-validation at lag h or lcv for leave-one-out cross-validation.
string, one of Wishart, smlnorm (log-Gaussian) or smnorm (Gaussian).
double, tolerance of optimization (root search)
lag step for consideration of observations, for the case criterion=lscv