Related Experiment Videos
Bump-hunting for the proficiency tester--searching for multimodality
Philip J Lowthian1, Michael Thompson
1School of Biological and Chemical Sciences, Birkbeck College, University of London, UK.
The Analyst
|November 15, 2002
Summary
Kernel density estimation and bootstrapping offer a robust method for analyzing proficiency test data. This approach helps identify multimodality and provides reliable estimates for assigned values and uncertainties.
Area of Science:
- Statistics
- Analytical Chemistry
- Proficiency Testing
Background:
- Histograms can present challenges in data approximation.
- Kernel density estimation (KDE) provides a smoother alternative.
- Proficiency testing data often requires robust statistical analysis.
Purpose of the Study:
- To apply KDE to proficiency test data for multimodality detection.
- To utilize bootstrapping for assessing the reliability of KDE.
- To establish a method for determining assigned values and uncertainties.
Main Methods:
- Kernel density estimation applied to z-scores from proficiency tests.
- Fitness-for-purpose criterion to determine smoothing parameters.
- Bootstrapping technique for data resampling and error estimation.
Main Results:
- KDE effectively identifies multimodality in z-score distributions.
- Bootstrapping confirms the ruggedness of the estimated kernel density.
- Standard errors of modes are accurately estimated.
Conclusions:
- KDE with bootstrapping is a suitable method for analyzing proficiency test data.
- The identified modes and their standard errors can serve as assigned values and uncertainties.
- This method enhances the reliability of proficiency testing evaluations.