# A comparison of location estimators for interlaboratory data contaminated with value and uncertainty outliers

- First Online:

- Received:
- Accepted:

DOI: 10.1007/s00769-008-0360-3

- Cite this article as:
- Duewer, D.L. Accred Qual Assur (2008) 13: 193. doi:10.1007/s00769-008-0360-3

## Abstract

While estimation of measurement uncertainty (MU) is increasingly acknowledged as an essential component of the chemical measurement process, there is little agreement on how best to use even nominally well-estimated MU. There are philosophical and practical issues involved in defining what is “best” for a given data set; however, there is remarkably little guidance on how well different MU-using estimators perform with imperfect data. This report characterizes the bias, efficiency, and robustness properties for several commonly used or recently proposed estimators of true location, *μ*, using “Monte Carlo” (MC) evaluation of “measurement” data sets drawn from well-defined distributions. These synthetic models address a number of issues pertinent to interlaboratory comparisons studies. While the MC results do not provide specific guidance on “which estimator is best” for any given set of real data, they do provide broad insight into the expected relative performance within broadly defined scenarios. Perhaps the broadest and most emphatic guidance from the present study is that (1) well-estimated measurement uncertainties can be used to improve the reliability of location determination and (2) some approaches to using measurement uncertainties are better than others. The traditional inverse squared uncertainty-weighted estimators perform well only in the absence of unrepresentative values (value outliers) or underestimated uncertainties (uncertainty outliers); even modest contamination by such outliers may result in relatively inaccurate estimates. In contrast, some inverse total variance-weighted-estimators and probability density function area-based estimators perform well for all scenarios evaluated, including underestimated uncertainties, extreme value outliers, and asymmetric contamination.

### Keywords

Consensus valueInterlaboratory comparisonsMeasurement uncertaintyMixture modelsMonte Carlo evaluationProbability density functionRobustnessWeighting function### Abbreviations

- CCQM
Comité Consultatif pour la Quantité de Matière

- ITV
Inverse total variance

- IU
^{2} Inverse squared uncertainty

- KC
Key comparison

- Lp
Least power

- MADe
Median absolute deviation from the median, expressed as a standard deviation

- MC
Monte Carlo

- MM
Mixture model

- MU
Measurement uncertainty

*n*Number of measurements in a data set

*n*_{BS}Number of bootstrap pseudo-data sets

*n*_{MC}Number of Monte Carlo simulation data sets

*N*(*μ*,*σ*)Normal (Gaussian) distribution having mean

*μ*and standard deviation*σ*- NMI
National Metrology Institute

*P*Level of confidence

- PC
Principal component

Probability density function

*U*_{I}[*ℓ*,*u*]Uniform (rectangular) distribution of integers having lower limit

*ℓ*and upper limit*u**U*_{R}[*ℓ*,*u*]Uniform (rectangular) distribution of real numbers having lower limit

*ℓ*and upper limit*u**s*Estimate of dispersion, expressed as a standard deviation

- \( s{\left( {\hat{\mu }} \right)} \)
Estimate of the variability of an estimator on replicate sampling of a population, expressed as a standard deviation

*u*(*x*_{i})Uncertainty component of the

*i*th measurement, expressed as a standard deviation*U*_{P}(*x*_{i})Uncertainty component of the

*i*th measurement, expressed as a coverage interval providing approximately*P*% level of confidence*w*_{i}Weight in a given calculation given to the

*i*th measurement- \( \ifmmode\expandafter\bar\else\expandafter\=\fi{x} \)
Arithmetic mean

*x*_{i}Value component of the

*i*th measurement*μ*True location of a population

- \( \hat{\mu } \)
Estimate of location

*σ*True dispersion of a population, expressed as a standard deviation