Partition features by comparing per-coordinate collapsed scores from two fits

partition_features_twofits(
  fitA,
  fitB,
  X = NULL,
  delta = 0,
  include_constant = TRUE
)

Arguments

fitA, fitB

csmooth_em objects (e.g., two different orderings / initializations).

X

Optional data matrix. If NULL, uses fitA$data (and assumes same X for both fits).

delta

Nonnegative margin. Assign to A only if (CA - CB) > delta; otherwise assign to B.

include_constant

Logical; passed to score_features_onefit().

Value

A list with - assign: length-d character vector in "A","B" - score_diff: length-d vector (CA - CB) - CA, CB: length-d vectors of per-feature scores