Evaluate biomarker performance — EvaluateBiomarkerPerformance • SciDataReportR

Evaluates a continuous or categorical biomarker against a binary or continuous outcome. Binary analyses use ordinary logistic regression. Models with separation, non-convergence, or aliased coefficients return stable unavailable metrics.

Usage

EvaluateBiomarkerPerformance(
  data,
  outcome_var,
  biomarker_var,
  covariates = NULL,
  PositiveLevel = NULL,
  OutcomeType = c("auto", "binary", "continuous"),
  ThresholdMethod = c("youden", "sensitivity", "specificity", "custom"),
  ThresholdValue = NULL,
  RawThresholdValue = NULL,
  ProbabilityThresholdValue = NULL,
  Validation = c("none", "bootstrap", "cross_validation"),
  BootstrapR = 500,
  CVFolds = 10,
  CIBootstrapR = 500,
  CILevel = 0.95,
  CalibrationGroups = 10,
  Seed = 123,
  Relabel = TRUE,
  codebook = NULL,
  Verbose = TRUE
)

Arguments

data: A data frame.
outcome_var: One outcome variable name.
biomarker_var: One biomarker variable name.
covariates: Optional covariate variable names.
PositiveLevel: Positive binary outcome level, or NULL to use the second observed level.
OutcomeType: One of "auto", "binary", or "continuous".
ThresholdMethod: One of "youden", "sensitivity", "specificity", or "custom".
ThresholdValue: Target sensitivity or specificity.
RawThresholdValue: Custom raw-biomarker threshold.
ProbabilityThresholdValue: Custom predicted-probability threshold.
Validation: One of "none", "bootstrap", or "cross_validation".
BootstrapR: Number of bootstrap optimism-correction resamples.
CVFolds: Number of cross-validation folds.
CIBootstrapR: Number of bootstrap confidence-interval resamples.
CILevel: Confidence level.
CalibrationGroups: Maximum grouped-calibration bins.
Seed: Random seed.
Relabel: Use codebook labels, then label attributes, for presentation.
codebook: Optional data frame with Variable and Label.
Verbose: Print positive-level information.

Value

A stable named list containing models, performance, thresholds, predictions, calibration, validation, plots, and metadata.

Examples

if (FALSE) EvaluateBiomarkerPerformance(df, "DiseaseCohort", "NfL", c("Age", "Sex")) # \dontrun{}