Compute Parameter Bounds for All Candidate Models — compute_model

Inspects the response range to classify the assay scale (high/medium/low), then builds adaptive bounds for a, b, c, d (and g for 5-param models).

Usage

compute_model_constraints(
  data,
  formulas,
  response_variable,
  independent_variable,
  is_log_response,
  fixed_a = NULL,
  verbose = FALSE
)

Arguments

data: Data frame with response and concentration columns.
formulas: Named list of formula objects.
response_variable: Character. Response column name.
independent_variable: Character. Concentration column name.
is_log_response: Logical. Is the response log10-transformed?
fixed_a: Numeric or NULL. If non-NULL, a is fixed and not bounded. If NULL, adaptive bounds for a are computed.
verbose: Logical.

Value

Named list of lists, each with lower and upper named vectors. Has attribute "constraint_profile".

Details

When fixed_a is NULL, the lower asymptote a is free and its bounds are set adaptively from the data. When fixed_a is non-NULL, a does not appear in the formulas (it's baked in), so no bounds are needed.