Medical AI Validation Example Dataset — medical_ai

A simulated dataset for demonstrating AI model validation methods. Contains AI predictions, human expert scores, clinical biomarkers, and gold-standard diagnosis for 200 patients.

Usage

medical_ai_data

Format

A data frame with 200 rows and 7 variables:

patient_id: Patient identifier (character)
AI_score: AI model predicted probability (0-1)
human_score: Human expert assessment score (0-1)
biomarker1: Clinical biomarker 1 (standardized)
biomarker2: Clinical biomarker 2 (standardized)
age: Patient age in years
diagnosis: Gold standard diagnosis (negative/positive)

Source

Simulated data generated for package demonstration

Details

This dataset was generated to demonstrate AI model validation techniques including cross-validation, calibration assessment, and model comparison. The AI_score is designed to have good but imperfect discrimination, making it suitable for demonstrating various validation metrics.

The outcome (diagnosis) was generated based on a logistic model with AI_score, biomarkers, and age as predictors, plus random noise to simulate real-world variability.

Examples

data(medical_ai_data)

# Basic cross-validation
if (FALSE) { # \dontrun{
aivalidation(
  data = medical_ai_data,
  predictorVars = c("AI_score", "biomarker1"),
  outcomeVar = "diagnosis",
  positiveLevel = "positive",
  crossValidation = "10-fold",
  showCrossValidation = TRUE
)
} # }

# Model comparison
if (FALSE) { # \dontrun{
aivalidation(
  data = medical_ai_data,
  predictorVars = c("AI_score", "human_score", "biomarker1", "biomarker2"),
  outcomeVar = "diagnosis",
  positiveLevel = "positive",
  compareModels = TRUE,
  delongTest = TRUE
)
} # }