Statistical Methods for Non-Detects - Guidance Document
Project Reference Number: 2008-03
Project Status: Complete
Led by: Greg Tamblyn, Impact Assessment Biologist, Ministry of Environment, Smithers
Carolyn Huston, Statistics and Actuarial Science, Simon Fraser UniversityFunder: Ministry of Environment, Grant Agreement
No standards exist in BC for scientists and technicians involved with analysing and interpreting water quality datasets that contain values below the method detection limit (MDL) of analytical equipment used by laboratories. Complicating matters, detection limits change over time and are different among laboratories. Hence researchers, consultants, development proponents and government staff use different methods for analyzing data. A common practice is to use substitution of values below the MDL with arbitrary (fabricated) numbers such as the MDL itself, ½ the MDL, or zero. Despite the prevalence of its use, numerous statistical studies have shown that such substitutions can lead to biased estimates and incorrect conclusions. With the current government focus on “results-based” monitoring, it is imperative to ensure that proper techniques are used to analyse data in order to provide accurate results.
The Ministry’s payment is to provide funds for developing material that will assist in standardizing statistical analysis protocols for datasets (particularly water quality data) containing values below method detection limits.
Objectives are to:
- develop statistical protocols to be used to properly analyze datasets containing non-detects and datasets containing inconsistent method detection limits;
- develop a detailed guidance document explaining how to conduct the necessary statistical analyses; and,
- distribute this guidance document to the scientific community and conduct training sessions in at least two places in the province.
Related Reports
Publication Date | Report Title | Authors |
---|---|---|
January 2009 | Guidelines for computing summary statistics for data-sets containing non-detects | C. Huston and E. Juarez-Colunga |