Paired / repeated-measures abundance test

WHAT: when samples pair up (pre/post, longitudinal, matched donors), compares WITHIN pair instead of between groups.

Usage

stats_paired_comparison(
  freq,
  pair_of,
  cond_of,
  reference = NULL,
  value_col = NULL,
  min_pairs = 3L
)

Arguments

freq

Population frequency table, one row per sample x population.

pair_of

named vector sample_id -> pairing unit (donor)

cond_of

named vector sample_id -> condition (timepoint / arm)

reference

The group every other group is compared against.

value_col

Column to test or plot instead of the default measure.

min_pairs

Minimum complete pairs before a paired test is attempted. Default 3L.

Details

WHY IT CANNOT BE INFERRED AND MUST BE DECLARED: nothing in an FCS file or a filename says two tubes came from the same donor at two timepoints. Treating paired samples as independent – which is what the baseline necessarily does, having no notion of pairing – discards the pairing's variance reduction AND violates the independence assumption the rank test rests on. So this runs only when –paired-column names a column, and it counts incomplete pairs rather than silently dropping them, because a pairing that quietly halves n is exactly the failure this function exists to prevent.

WHY WILCOXON SIGNED-RANK / FRIEDMAN: the paired counterparts of the rank tests used everywhere else here. Friedman generalises to 3+ conditions and is the conventional choice for repeated-measures frequency data at this scale.