Year
1994
Abstract
Before data is used in any statistical evaluation, it should be subjected to a thorough data qualification to ensure its integrity. Data qualification efforts may vary in scope but should all contain certain basic elements. Outliers should be eliminated so that any estimates obtained from the data will not be overly biased. Plots may be generated to uncover potential outliers, and these observations can be confirmed as being anomalous through the use of statistical testing. Non-stationarity of the data (i.e., time trends) may also be initially identified through the use of plots and confirmed by the results of statistical analyses, e.g., the use of the Kendall's tau statistic. When the dataset has been qualified, replicates and duplicates should be averaged to create a pseudo-independence between data points. An algorithm based on the groundwater data from the Savannah River Site has been used by the Applied Statistics Group at Westinghouse Savannah River Company for this.