Concordance between the raw-percentage test and the CLR test — compositional

WHAT THE READER GETS: for every population x comparison, whether the conclusion survives removing the compositional constraint, with a one-word verdict naming which of the four cases it falls into.

Usage

compositional_concordance(
  raw_stats,
  clr_stats,
  alpha = 0.05,
  p_col = c("p_value", "p_adj_BH")
)

Arguments

raw_stats: The raw stats.
clr_stats: The clr stats.
alpha: Significance threshold. Default 0.05.
p_col: Which p-value column to compare on. Default c("p_value", "p_adj_BH").

Details

WHY A TABLE AND NOT A LOG LINE: "significant on raw but not on CLR" is a per-population fact and populations behave differently. A single summary sentence would let a reader carry a global impression onto the one population where it does not hold.