Toward improved analysis of concentration data: Embracing nondetects

Area of Science:

• Environmental Science
• Toxicology
• Statistics

Background:

• Routine statistical methods are inadequate for environmental concentration data containing nondetects (left-censored observations).
• Existing techniques for handling censored data are often overlooked or improperly applied.
• Regulatory guidance on censored data analysis may be outdated, leading to inconsistent practices.

Purpose of the Study:

• To address the challenges posed by nondetects in environmental data analysis.
• To provide researchers with theoretical and computational recommendations for analyzing censored concentration data.
• To improve the interpretation of environmental monitoring and risk assessment data.

Main Methods:

• Review of existing literature on censored data analysis in environmental science.
• Synthesis of findings related to estimation, hypothesis testing, and regression analysis.
• Development of recommendations within both frequentist and Bayesian frameworks.

Main Results:

• Common practice of substituting nondetects with arbitrary constants is a bias-prone approach.
• Proper statistical adjustment for nondetects is crucial for accurate environmental data interpretation.
• Discrepancies exist in current policies and guidance for managing censored data.

Conclusions:

• Accurate statistical analysis of censored environmental data is essential for reliable decision-making.
• This article provides practical guidance for researchers and practitioners dealing with nondetects.
• Adopting recommended methods will enhance the validity of environmental risk assessments.