Initialize ordering for csmoothEM (diagonal-variance version) — initialize_ordering

Same ordering logic as initialize_ordering(), but returns csmooth-style parameters with diagonal variances sigma2.

initialize_ordering_csmooth(
  X,
  K,
  method = c("PCA", "fiedler", "pcurve", "tSNE", "random", "isomap"),
  discretization = c("equal", "quantile", "kmeans"),
  modelName = c("homoskedastic", "heteroskedastic"),
  nugget = 0,
  eps = 1e-12,
  ...
)

Arguments

X: Numeric matrix (n x d).
K: Integer >= 2; number of mixture components.
method: One of "PCA","fiedler","pcurve","tSNE","random", "isomap".
discretization: One of "equal","quantile","kmeans".
modelName: Either "homoskedastic" or "heteroskedastic".
nugget: Nonnegative scalar added to variance estimates.
eps: Small positive floor for pi and sigma2.
...: Extra arguments passed to the ordering method (PCA/tSNE/pcurve/fiedler).

Value

A list with fields:

params: list(pi, mu, sigma2)
keep_idx: indices kept after NA handling
ordering: ordering metadata including score t and method name