Build a distributional-data (dd) object

Constructor for a dd object: a B-spline fd representation of a log-density together with its normalizing constant (so the density is exp(W(x)) / constant).

Usage

dd(
  sample = NULL,
  full_sample = sample,
  basis = fda::create.bspline.basis(rangeval = rangeval, nbasis = nbasis, norder =
    norder, breaks = breaks),
  knots_pos = "quantiles",
  breaks = if (is.numeric(sample)) {
     if (knots_pos == "quantiles") {
        
    quantile(full_sample, seq(0, 1, length.out = nbasis - norder + 2))
     }
     else
    {
         quantile(range(full_sample), seq(0, 1, length.out = nbasis - norder + 2))

        }
 } else {
     NULL
 },
  rangeval = if (is.numeric(sample)) range(breaks) else c(0, 1),
  nbasis = 12,
  norder = 5,
  lambda = 0,
  clr = fda::fd(basisobj = basis),
  constant = NULL,
  normalize = TRUE,
  ...
)

Arguments

sample: Optional numeric sample used to estimate the density.
full_sample: Pooled reference sample for knot placement. Defaults to sample.
basis: A basisfd object. Defaults to a B-spline basis built from rangeval, nbasis, norder, breaks.
knots_pos: How to derive knots from full_sample ("quantiles" by default).
breaks: Numeric vector of knot positions. By default chosen from full_sample according to knots_pos; NULL for non-numeric input.
rangeval: Numeric pair giving the range over which the density lives. Defaults to range(breaks) for numeric input, c(0, 1) otherwise.
nbasis: Number of basis functions. Default 12.
norder: B-spline order. Default 5.
lambda: Penalty parameter forwarded to density_mpl().
clr: Pre-computed clr-transformed fd (skip density estimation if supplied).
constant: Optional pre-computed normalizing constant.
normalize: Logical; if TRUE (default) compute the normalizing constant via normalize().
...: Further arguments forwarded to density_mpl().

Value

A dd object: an fd carrying constant and (optionally) sample.