Classical statistical methods that use the sample mean and standard deviation, under the assumption that the data follow a Gaussian distribution, are often applied to measurement intercomparisons. Consequently, spurious measurements, or outliers, lead to inaccuracies in the estimated mean and standard deviation, thus biasing the results of measurement comparisons. This paper describes the application of robust statistical methods for analyzing measurement results from an intercomparison containing spurious results. Robust techniques eliminate the need to subjectively separate the data into "good" and "bad" before calculating the statistics. We apply these techniques to results obtained from a measurement intercomparison of vector network analyzers.
展开▼
