Skip to contents

Scores the pools (pool_scores() per pool x region), applies the fold-0 pooled thresholds per class and returns the policy's pool-level rows. With loo the pools are scored as pool_scores() describes; use score = "s_rob" (thresholded by pooled_thresholds() with loo = TRUE) so a call must survive dropping its most influential cell.

Usage

call_pools(
  d,
  pools,
  thr,
  policy = call_policy("normal"),
  states = caller_states("normal"),
  regions = NULL,
  score = "s_comb",
  loo = score %in% c("s_rob", "s_loo"),
  scale = NULL
)

Arguments

d

rows scored with keep_rows = TRUE

pools

pools table (pool, cell)

thr

pooled_thresholds() output

policy

call_policy()

states

state set used for scoring

regions

optional regions table (coordinates)

score

pooled score

loo

leave-one-out scores (needed for score = "s_rob")

scale

NULL or the pool_scale() table the thresholds were computed with (attribute pool_scale of pooled_thresholds() with standardize = TRUE). A region of a class without null pools (named event regions) is standardised with the class medians of its sd_class when thr has borrowed rows

Value

calls table (level = "pool", cell = pool id) plus n (cells in the pool), s_comb and, with loo, s_loo; with scale, s_comb_llr (unstandardised best LLR)