Full Preprocessing Pipeline for Standard Curve Data (mask-aware)

Applies concentration computation, prozone correction, blank handling, and optional log10 response transform in the canonical order.

Usage

preprocess_standards(
  data,
  antigen_settings,
  response_variable,
  independent_variable,
  is_log_response,
  blank_data = NULL,
  blank_option = "ignored",
  is_log_independent = TRUE,
  apply_prozone = TRUE,
  include_col = "included",
  verbose = FALSE
)

Arguments

data

Data frame of standards with a dilution column.

antigen_settings

Named list with standard_curve_concentration.

response_variable

Character. Response column name.

independent_variable

Character. Concentration column name.

is_log_response

Logical. Log10-transform the response?

blank_data

Data frame of blanks, or NULL.

blank_option

Character. Blank handling method.

is_log_independent

Logical. Log10-transform concentration?

apply_prozone

Logical. Apply prozone correction?

include_col

Character. Name of the logical include/mask column on data and blank_data. Absent on a frame = all its rows are included.

verbose

Logical.

Value

A named list:

data

All standard rows, transformed, carrying include_col, the (log10) concentration, the model-space response, and assay_response_raw.

blanks

All blank rows, transformed onto the standards' response floor, carrying include_col, the model-space response, and assay_response_raw; NULL when no blank_data was supplied.

antigen_fit_options

Record of the options used.

derived_stats

The set-level statistics computed from the included points: blank_geomean, prozone_peak_response, prozone_logc_at_peak, response_floor, min_included_concentration.

Details

Mask-aware contract. An include_col logical column (default "included", TRUE = used in the fit, FALSE = masked) may be present on data and blank_data. Every set-level statistic — the prozone peak, the blank geometric mean, the adaptive log floor, and the minimum-concentration anchor — is computed from the included rows only. The resulting transforms are then applied to all rows, so masked points land on the same axes as the fitted points without ever influencing them. If the column is absent, every row is treated as included and the output is identical to the pre-mask behaviour (backward compatible).

The function does not drop masked rows; downstream fitters are expected to receive only the included subset (e.g. pp$data[pp$data$included, ]), which keeps the fit byte-identical to a fit that never saw the masked rows.

Both frames retain a pristine assay_response_raw column (the response before prozone/blank/log), so callers can persist the raw and model-space responses side by side.

Blanks are never subtracted automatically. With the default blank_option = "ignored" the standard responses are left untouched; the returned blanks frame is transformed for display/persistence only and is not subtracted from the standards. Subtraction happens only when the caller explicitly selects "subtracted", "subtracted_3x", or "subtracted_10x" (which subtract 1x/3x/10x the included-blank geometric mean), or adds the blank mean as a point via "included". The returned blanks are always the raw and model-space blank responses, never a subtracted quantity.