Calculate the treatment-to-control melanization coefficient — melanization

Computes MC = mean treatment plate MI / mean control plate MI. Each plate MI is the arithmetic mean of its target colony MI_mean values. Plates are equally weighted, irrespective of pixel counts or numbers of colonies. This is a ratio of descriptive image-darkness indices, not a ratio of melanin concentrations. MC > 1 indicates a darker treatment mean; MC < 1 indicates a lighter treatment mean. No significance test or CI is implied.

Usage

melanization_coefficient(
  colonies,
  control = "control",
  group = "treatment",
  plate = "plate_id",
  response = "MI_mean",
  by = NULL,
  exclude_ids = c("centre", "center"),
  na_action = c("error", "omit"),
  set = "set_id",
  require_qc = FALSE
)

Arguments

colonies: Colony data frame returned by analyze_plates(), or such a batch result. Must contain group, plate, colony_id and response columns. Only target Zymoseptoria rows should be supplied. The standard central partner ID is excluded automatically; specify other partner IDs below.
control: Exact group label for the no-interaction control.
group: Treatment column. Every non-control level is compared separately.
plate: Column identifying an independent plate uniquely.
response: Colony-level index column, default MI_mean (range 0 to 100).
by: Optional matching columns, e.g. c("zymo_strain", "day", "medium"). A control must exist within each combination. Do not include partner identity when controls do not share it. Missing matching values are errors.
exclude_ids: Colony identifiers excluded before calculation. Default excludes centre and center, used for the central interacting fungus.
na_action: Either error (default) for undetected/nonfinite target colonies or omit, which warns and reports excluded counts. A plate with no eligible target measurements remains unavailable, not silently dropped.
set: Matching-set column, required. Controls are never pooled across sets.
require_qc: Require every included plate to have logical qc_pass=TRUE. Default FALSE warns for unconfirmed recorded QC and continues without approving or excluding plates. Plain tables without review flags are reported as not_recorded. Finalization warns if review is unrecorded.

Value

List with coefficients (one row per contrast and matching stratum), plate_means, and settings. Undefined ratios have NA and an explicit status. Counts distinguish measured target colonies, excluded target rows and plates. One plate per group permits a descriptive ratio, not biological inference.

Examples

d <- data.frame(plate_id = rep(c("C1", "T1"), each = 4),
                set_id = "set1",
                treatment = rep(c("control", "fungus"), each = 4),
                colony_id = rep(c("TL", "TR", "BR", "BL"), 2),
                MI_mean = c(40, 42, 38, 40, 60, 62, 58, 60))
melanization_coefficient(d)$coefficients  # 60 / 40 = 1.5
#>   set_id treatment control MI_treatment_mean MI_control_mean
#> 1   set1    fungus control                60              40
#>   melanization_coefficient delta_MI n_treatment_plates n_control_plates
#> 1                      1.5       20                  1                1
#>   qc_review_status  control_replication n_treatment_colonies n_control_colonies
#> 1      unconfirmed single_control_plate                    4                  4
#>   n_excluded_target_colonies status                         interpretation
#> 1                          0     ok Higher treatment image-darkness index.