Pairwise Log-Ratio Variance Feature Filter — codaFS_plr

Scores pairwise log-ratios by combining their across-sample variance with their association to the binary outcome, then aggregates the highest-scoring ratios back to individual features by weighted frequency.

Usage

codaFS_plr_variance(
  X,
  y,
  n_features = NULL,
  positive = NULL,
  input_scale = c("log", "raw"),
  top_pairs = NULL,
  pair_multiplier = 5L,
  max_pair_features = 200L,
  prescreen_metric = c("clr_effect", "variance")
)

Arguments

X: Numeric matrix with samples in rows and features in columns.
y: Binary outcome vector.
n_features: Optional target panel size. Used only to derive the default number of pairwise log-ratios retained during aggregation.
positive: Optional positive-class label for non-numeric outcomes.
input_scale: Either `"log"` (default) or `"raw"`.
top_pairs: Number of top pairwise log-ratios to aggregate. If `NULL`, this is set to `pair_multiplier * n_features` or a conservative default.
pair_multiplier: Multiplier used when `top_pairs` is `NULL`.
max_pair_features: Maximum number of features retained before computing all pairwise log-ratios. Prescreening uses training-fold-only CLR effects.
prescreen_metric: Either `"clr_effect"` or `"variance"`.

Value

A list with per-feature `scores`, `selected_features`, and the aggregated top pairwise log-ratio table.