Log-normal and Weighted Exponential Imputation: Advanced Techniques for Handling Values Below the Detection Limit (DL)

This study addresses the challenge of handling values below the detection limit (DL) in environmental and toxicological data analysis. Traditional imputation methods, such as substituting censored values with constants, often distort data distributions and lead to unreliable statistical outcomes. To improve accuracy, the study explores two advanced imputation techniques: log-normal and weighted exponential imputation, which better reflect the skewed patterns typical of environmental contamination and exposure levels. A third approach, Alpha-Based Hybrid Imputation, is introduced, combining the strengths of both log-normal and weighted exponential methods. This hybrid technique enhances the precision of imputed values, particularly for datasets with values clustered near the DL. Python code for all three methods is presented, and the imputation process and validation are further documented in separate Excel files for 32 CEMET metals collected 2 hours after intravenous application of standard doses of Glutathione, Na-Ca EDTA, and DMPS.